Alphabet Inc. (GOOGL) 2026 Q3 法說會逐字稿

Welcome, everyone. Thank you for standing by for the Alphabet third-quarter 2026 earnings conference call.

歡迎各位。感謝各位撥冗參加 Alphabet 2026 年第三季財報電話會議。

(Operator Instructions)

(接線員指示)

I would now like to hand the conference over to your speaker today, Jim Friedland, Head of Investor Relations. Please go ahead.

現在我想將會議交給今天的講者——投資人關係主管 Jim Friedland。請開始。

Thank you. Good afternoon, everyone, and welcome to Alphabet's third quarter 2026 earnings conference call. With us today are Sundar Pichai; Philipp Schindler; Thomas Kurian; and Anat Ashkenazi. Now I'll quickly cover the safe harbor.

謝謝。各位下午好，歡迎參加 Alphabet 2026 年第三季財報電話會議。今天與會者包括 Sundar Pichai、Philipp Schindler、Thomas Kurian，以及 Anat Ashkenazi。接下來我將快速說明安全港聲明。

Some of the statements that we make today regarding our business, operations, and financial performance may be considered forward-looking. Such statements are based on current expectations and assumptions that are subject to a number of risks and uncertainties. Actual results could differ materially. Please refer to our forms 10-K and 10-Q, including the risk factors. We undertake no obligation to update any forward-looking statement.

我們今天就業務、營運與財務表現所做的部分陳述，可能被視為前瞻性陳述。此類陳述係基於目前的預期與假設，並受多項風險與不確定性影響。實際結果可能有重大差異。請參閱我們的 10-K 與 10-Q 表格申報文件（包括風險因素）。我們不承擔更新任何前瞻性陳述之義務。

During this call, we will present both GAAP and non-GAAP financial measures. A reconciliation of non-GAAP to GAAP measures is included in today's earnings press release, which is distributed and available to the public through our investor relations website located at abc.xyz/investor. Our comments will be on year-over-year comparisons unless we state otherwise.

在本次電話會議中，我們將同時呈現 GAAP 與非 GAAP 財務衡量指標。非 GAAP 與 GAAP 指標之調節表已納入今日發布的財報新聞稿，並已透過我們位於 abc.xyz/investor 的投資人關係網站對外發布且可供大眾查閱。我們的評論除非另有說明，將以年對年比較為主。

And now, I'll turn the call over to Sundar.

接下來我把電話交給 Sundar。

Thanks, Jim, and hello, everyone. Q3 was a strong quarter for Alphabet. Our deep full-stack approach is driving momentum across the business with many opportunities ahead.

謝謝你，Jim，各位好。第三季對 Alphabet 而言是表現強勁的一季。我們深度的全堆疊策略正推動整體業務動能，未來仍有許多機會。

Let me give you a quick snapshot of the quarter. Alphabet revenues grew 25% year over year. We are now at a $0.5 trillion annual run rate. That's remarkable growth. Search has strong momentum. Our AI investments are driving increased usage and overall query growth, including commercial queries.

我先快速概述本季表現。Alphabet 營收年增 25%。我們目前的年化營收運行率已達 0.5 兆美元。這是非常驚人的成長。搜尋業務動能強勁。我們在 AI 的投資正帶動使用量提升與整體查詢成長，包括商業查詢。

In fact, our year-over-year query growth rate is the highest it's been in the last five years. Google Cloud revenues grew 103%. Our AI infrastructure and solutions are the major drivers for cloud growth, and our customers see us as the leading provider here to those huge demand. We are working hard to bring more capacity online to meet this long-term demand.

事實上，我們的查詢年增率是過去五年來最高。Google Cloud 營收成長 103%。我們的 AI 基礎設施與解決方案是雲端成長的主要驅動力，客戶也將我們視為能滿足龐大需求的領先供應商。我們正努力擴增上線產能，以滿足這項長期需求。

At the same time, we are seeing strong demand for our TPU systems across a wide range of customers, and we expect that to continue as we scale. I'm excited to welcome Thomas to the call, and he'll share more about Cloud's Momentum.

同時，我們也看到各類客戶對 TPU 系統有強勁需求，並預期隨著我們擴大規模，這股需求將持續。我很高興 Thomas 也加入今天的電話會議，他將分享更多關於雲端業務動能的內容。

On our AI models, customers are using our models more than ever. Our first-party model APIs are now processing nearly 32 billion tokens per minute. 40% more than last quarter and more than four times what we processed a year ago.

在 AI 模型方面，客戶使用我們模型的程度比以往任何時候都更高。我們第一方模型 API 目前每分鐘處理的 token 量接近 320 億。較上季增加 40%，也比一年前處理量高出四倍以上。

And with Gemini 4 Argon and now Carbon, which we announced today, we are excited to be leading at the frontier. The Gemini app is now at 1 billion users and is the fastest growing product in Google's history. Lastly, we have surpassed 150 million YouTube music and premium subscribers globally, including trials.

隨著 Gemini 4 Argon，以及我們今天宣布的 Carbon，我們很興奮能在前沿領域保持領先。Gemini 應用程式目前已有 10 億使用者，並且是 Google 歷史上成長最快的產品。最後，我們在全球的 YouTube Music 與 Premium 訂閱用戶（含試用）已突破 1.5 億。

Today, I'll talk more about our AI progress, our global consumer platforms led by Search and YouTube, then our long-term bets. Starting with AI, highlighted by our Gemini 4 series of frontier models, we are seeing strong demand and usage for Argon, which we launched two weeks ago.

今天我將進一步談談我們在 AI 的進展、以搜尋與 YouTube 為首的全球消費者平台，以及我們的長期押注。先從 AI 談起，以 Gemini 4 系列前沿模型為亮點，我們看到兩週前推出的 Argon 需求與使用量都很強勁。

In fact, on our developer platforms, usage in the first week ramped up 3.5 times faster than Gemini 3, even with capacity constraints. Argon is proving to do really well in areas like finance, legal, cybersecurity and science.

事實上，在我們的開發者平台上，即使受限於產能，第一週的使用量爬升速度仍比 Gemini 3 快了 3.5 倍。Argon 在金融、法律、資安與科學等領域表現非常出色。

Developers and cloud customers are already using it for things like untangling legacy code bases, accelerating scientific research, robotics, discovering product improvements and domain-specific tasks requiring in-depth reasoning.

開發者與雲端客戶已經用它來處理例如梳理既有遺留程式碼庫、加速科學研究、機器人技術、發掘產品改進，以及需要深度推理的特定領域任務。

And at Google, just in the last few weeks, teams across Chrome, Ads, and Search have used Argon to write and optimize code much faster, shipping 35% more code with 30% fewer errors. Argon is just the start of the Gemini 4 series.

而在 Google，僅在過去幾週內，Chrome、Ads 與 Search 等團隊已使用 Argon 更快速地撰寫與最佳化程式碼，在錯誤減少 30% 的同時，交付的程式碼量增加 35%。Argon 只是 Gemini 4 系列的起點。

Today, we announced Carbon. It's our strongest frontier model with state-of-the-art cybersecurity and long horizon capabilities. It ranks number one on the recently updated Artificial Analysis Intelligence Index at 59. Carbon outperforms Argon across almost every domain like full stack web development, multimodal understanding, complex knowledge work, and scientific reasoning. We are rolling Carbon out first to a subset of trusted cyber defenders before we expand access and it's being tested by the US government.

今天我們宣布了 Carbon。它是我們最強的前沿模型，具備最先進的資安能力與長期視野能力。在近期更新的 Artificial Analysis Intelligence Index 中以 59 分排名第一。Carbon 在幾乎所有領域都優於 Argon，例如全堆疊網頁開發、多模態理解、複雜知識工作與科學推理。我們將先把 Carbon 推出給一部分受信任的資安防禦者，之後再擴大開放；目前也正在接受美國政府測試。

Beyond our Gemini 4 series, yesterday we announced Genie 4, our most advanced world model yet that simulates safe, realistic, interactive worlds in real time. With native audio, it is an important building block for training robots and agents. This rapid progress underscores our incredible model momentum.

除了 Gemini 4 系列之外，我們昨天宣布了 Genie 4，這是我們迄今最先進的世界模型，可即時模擬安全、逼真且可互動的世界。它具備原生音訊能力，是訓練機器人與代理（agents）的重要基礎模組。這樣的快速進展凸顯了我們驚人的模型動能。

We are at the point where AI is helping us increase our training and launch velocity. I would expect the next 12 months to see continued acceleration because of RSI. We have more models coming soon, including more efficient versions to help developers optimize for speed and cost and to deploy across our billions of users.

我們已經來到 AI 正在幫助我們提升訓練與產品推出速度的階段。我預期未來 12 個月會因 RSI 而持續加速。我們很快還會推出更多模型，包括更高效率的版本，協助開發者在速度與成本之間做最佳化，並部署到我們數十億使用者之上。

Let's turn to Search, which is bringing the benefits of AI to more people than any other product in the world. Just like our shift to mobile over a decade ago, we are focused on building the best search user experience followed by monetization. We are building one seamless experience across AI overviews, AI mode, and traditional search. And people are using our AI features more, searching in entirely new ways. Their queries are longer, personalized, and multimodal.

接著談搜尋，它把 AI 的好處帶給的人數超過全球任何其他產品。就像十多年前我們轉向行動端一樣，我們專注於先打造最佳的搜尋使用者體驗，再進行變現。我們正在打造一個無縫體驗，整合 AI 摘要（AI overviews）、AI 模式（AI mode）與傳統搜尋。人們也更常使用我們的 AI 功能，以全新的方式進行搜尋。他們的查詢更長、更個人化，且是多模態的。

Queries in AI mode are on average three times longer than traditional searches, and one in five AI mode searches globally use voice, images or documents. We have now crossed over 50 billion visual searches per month. One area I'm excited about is helping people get things done, be it booking reservations, filling in forms, scheduling appointments, or making purchases.

AI 模式下的查詢平均比傳統搜尋長三倍，而全球每五次 AI 模式搜尋中就有一次使用語音、圖片或文件。我們目前每月的視覺搜尋量已突破 500 億次。我特別期待的一個方向，是協助人們把事情完成，無論是預訂、填寫表單、安排約會或進行購買。

As AI becomes more personal and starts taking action for people and businesses, trust and security matter more than ever. People already put their trust in Google to get things done in their daily lives, and their leadership here remains an important differentiator. We focus first on getting these agentic experiences right with the Gemini Agent for Enterprise because the bar for security, safety, and governance is extremely high.

隨著 AI 變得更個人化，並開始為個人與企業採取行動，信任與安全就比以往任何時候都更重要。人們早已信任 Google 協助他們處理日常生活中的各種事務，而我們在這方面的領先地位仍是重要的差異化優勢。我們會先在企業用的 Gemini Agent 上把這些代理式（agentic）體驗做到位，因為在安全性、保障與治理方面的門檻極高。

As a next step, we are bringing these features to Search and also, of course, to our billion users on the Gemini app. People are already using the Gemini app to do more than ask questions. They're using it to do things, including complex multi-step agent inquiries.

下一步，我們將把這些功能帶到搜尋中，當然也會帶給 Gemini 應用程式上的 10 億使用者。人們已經不只是用 Gemini 應用程式來提問。他們也用它來完成事情，包括複雜的多步驟代理式查詢。

Our Gemini 4 series has been a really big step forward here and has already been handling these complex tasks really well. Agents will operate as the unifying layer across our products, so you will see us extend these agentic experiences beyond Search in the Gemini app to Chrome, Android, and our workspace products as well.

我們的 Gemini 4 系列在這方面是一次非常重大的躍進，並且已經能夠非常出色地處理這些複雜任務。代理（Agents）將作為我們各項產品之間的統一層，因此你會看到我們把這些代理式體驗從 Gemini 應用程式中的搜尋延伸到 Chrome、Android，以及我們的 Workspace 產品。

On to YouTube, which continues to lead in the living room as the number one streaming platform in the US in terms of watch time. It's also seeing good subscriber growth, as I mentioned. At [Made] on YouTube last month, we announced more than 30 products and features for creators and viewers. One great example is custom feeds where you can simply pick a topic you love and get a tailored video home feed based on that topic. Phillip will say more on YouTube.

接著談 YouTube。以觀看時長計算，YouTube 持續在客廳場景領先，並且是美國第一大串流平台。如我先前提到的，它的訂閱用戶成長也相當不錯。在上個月的 YouTube「Made」活動上，我們為創作者與觀眾宣布了超過 30 項產品與功能。其中一個很好的例子是自訂動態（custom feeds）：你只要選擇你喜愛的主題，就能依該主題獲得量身打造的影片首頁動態。Phillip 會再補充更多關於 YouTube 的內容。

On to our long-term bets. Waymo is on a great trajectory. We have now driven more than 300 million fully autonomous miles. Waymo now operates in 15 major US cities and just recently opened up service to the public in San Antonio, San Diego, and Tampa with planned expansions into Canada, Germany, Singapore, and Spain.

接下來談我們的長期押注。Waymo 正處於非常好的發展軌道上。我們目前已累計行駛超過 3 億英里的全自動駕駛里程。Waymo 現在在美國 15 個主要城市營運，並且最近已在聖安東尼奧、聖地牙哥與坦帕向大眾開放服務，同時規劃擴展至加拿大、德國、新加坡與西班牙。

We're also seeing strong progress across our other long-term initiatives. Isomorphic is using Argon to support drug design research. We demonstrated magic state cultivation for the first time in quantum computing and launched our TPUs into space with Project Suncatcher. So it's been a busy quarter, and there's much more to come. I want to thank our employees for their work this quarter.

我們在其他長期計畫上也看到強勁進展。Isomorphic 正在使用 Argon 來支援藥物設計研究。我們在量子運算領域首次展示了魔態培育（magic state cultivation），並透過 Project Suncatcher 將我們的 TPU 發射到太空。因此這是一個忙碌的季度，而且後續還有更多值得期待。我要感謝我們的員工在本季度的付出。

And over to you, Phillip.

接下來交給你，Phillip。

Thanks, Sundar, and hello, everyone. I'll highlight our momentum across Search, YouTube, and partnerships. We continue to accelerate the deployment of Gemini models across our entire ads infrastructure to boost performance in the three areas I've highlighted before: ads quality, advertiser tools, and AI user experiences.

謝謝你，Sundar，大家好。我將重點說明我們在搜尋、YouTube 與合作夥伴關係方面的動能。我們持續加速在整個廣告基礎架構中部署 Gemini 模型，以提升我先前提到的三個面向的表現：廣告品質、廣告主工具，以及 AI 使用者體驗。

First, ads quality. The integration of Gemini models into our ads infrastructure is improving how we decode user intent and drive ads relevance. For example, in Q3, we made the biggest update to product listing ads in a decade, adding new AI capabilities to help users make more confident purchase decisions. Already, we've seen satisfaction increase by 19% and dissatisfaction drop more than half.

第一，廣告品質。將 Gemini 模型整合進我們的廣告基礎架構，正在改善我們解讀使用者意圖並提升廣告相關性的方式。例如在第三季，我們對商品列表廣告（product listing ads）進行了十年來最大的一次更新，加入新的 AI 能力，協助使用者更有信心地做出購買決策。目前我們已看到滿意度提升 19%，不滿意度下降超過一半。

We're also seeing AI models enable better prediction of clicks and conversions, stronger overall performance, and deeper personalization of ads. For instance, through several innovations and personalization modeling, we delivered 18% incremental advertiser value and an 8% increase in user sentiment on Discover Ads in 2026. Overall, AI is accelerating our ads development and drives the majority of our product launches.

我們也看到 AI 模型能更好地預測點擊與轉換、帶來更強的整體表現，並讓廣告個人化更深入。例如，透過多項創新與個人化建模，我們在 2026 年為 Discover Ads 帶來 18% 的增量廣告主價值，並使使用者情緒指標提升 8%。整體而言，AI 正在加速我們的廣告開發，並推動我們多數的產品發布。

Second, advertiser tools. We're ensuring every advertiser, from global brands to small businesses can seamlessly participate in our AI services. AI Max and Performance Max continue to see strong adoption. To give marketers even more control, we're now introducing a way to test different budgets and ROI targets across multiple search campaigns in a single AI Max A/B test. To drive better performance, we integrated our AI creative tools directly into AI Max.

第二，廣告主工具。我們確保每一位廣告主，從全球品牌到中小企業，都能無縫參與我們的 AI 服務。AI Max 與 Performance Max 持續獲得強勁採用。為了讓行銷人員擁有更多控制權，我們現在推出一種方式：在單一 AI Max A/B 測試中，跨多個搜尋活動測試不同預算與 ROI 目標。為了帶動更好的成效，我們也將 AI 創意工具直接整合進 AI Max。

In Q3 alone, Gemini models helped advertisers generate over 1.5 billion campaign assets, up over 7x year over year. For example, by using AI Max to capture long-form and multilingual queries, Telstra saw a 52% increase in conversions and a 36% reduction in cost per order, showing up in key moments however their customers search.

僅在第三季，Gemini 模型就協助廣告主生成超過 15 億個活動素材，年增超過 7 倍。例如，透過使用 AI Max 來涵蓋長尾與多語言查詢，Telstra 的轉換提升 52%，每筆訂單成本降低 36%，能在客戶以各種方式搜尋時於關鍵時刻曝光。

Third, AI user experiences. Sundar mentioned the distinct behavioral changes we're seeing as users ask entirely new categories of questions. We think of search holistically, a unified product spanning traditional results and AI-powered search.

第三，AI 使用者體驗。Sundar 提到，我們看到使用者在行為上出現明顯變化，因為他們開始提出全新類別的問題。我們以整體角度看待搜尋：這是一個統一的產品，涵蓋傳統結果與 AI 驅動的搜尋。

On AI mode, many of these queries are up-of-funnel queries, allowing us to capture new high-value commercial intent. In addition, the majority of AI mode queries are incremental, creating more monetization opportunities over time.

在 AI 模式中，許多查詢屬於漏斗上層（up-of-funnel）問題，使我們能捕捉新的高價值商業意圖。此外，多數 AI 模式查詢屬於增量查詢，隨時間推進將創造更多變現機會。

Additionally, our user research shows that people are comfortable with ads on these surfaces and find them highly relevant. Click-through rates are already similar between sponsored and organic product list items within AI mode. Based on these collective results, we're now scaling high-quality ads within AI overviews in AI mode to more users and queries. This includes taking our monetization footprint global for the first time.

另外，我們的使用者研究顯示，人們對在這些介面上出現廣告感到自在，並認為其高度相關。在 AI 模式中，贊助與自然的商品列表項目，其點擊率已經相近。基於這些綜合結果，我們現在正把 AI 模式中的 AI 概覽（AI overviews）內的高品質廣告擴大到更多使用者與查詢。這也包括我們首次將變現版圖推向全球。

Today, we're announcing the expansion of ads in AI mode into international markets alongside multi-turn ads in AI mode in the US. As always, our approach is deliberate, prioritizing the user experience and engagement first, and then scaling the formats that drive proven ROI for advertisers. That discipline is what gives us a strong, durable foundation for the future.

今天，我們宣布在美國推出 AI 模式中的多輪（multi-turn）廣告，同時把 AI 模式中的廣告擴展到國際市場。一如既往，我們的做法是審慎推進：先以使用者體驗與互動為優先，接著再擴大那些能為廣告主帶來已驗證 ROI 的格式。正是這份紀律，讓我們為未來建立強健且可長期維持的基礎。

As Sundar mentioned, we're in the early stages of agentic experiences across our services and excited about the commercial opportunities ahead. The Universal Commerce Protocol, UCP, is being rapidly adopted by retailers. Today, UCP enables direct checkout for hundreds of thousands of brands across Google, including Chewy, Nike, Tapestry, and many more.

如 Sundar 所提到的，我們正處於在各項服務中推動代理式體驗的早期階段，並對未來的商業機會感到振奮。通用商務協定（Universal Commerce Protocol，UCP）正被零售商快速採用。目前，UCP 已在 Google 各項服務中為數十萬個品牌提供直接結帳能力，包括 Chewy、Nike、Tapestry 等眾多品牌。

We're combining this with the agent payments protocol as we build for the future of commerce. Our focus spans two fronts, unlocking agentic capabilities that act on behalf of the consumer and designing practical tools that give shoppers confidence and control.

在打造未來商務的同時，我們也把這與代理支付協定（agent payments protocol）結合。我們的重點涵蓋兩個面向：一是解鎖可代表消費者採取行動的代理式能力；二是設計實用工具，讓購物者更有信心並保有掌控權。

We launched Universal Cart this quarter to support that second track, mirroring how people actually shop today across Google. This allows retailers to maintain their relationships with customers while helping to capture sales and avoid abandoned carts.

本季度我們推出 Universal Cart 以支援第二條路線，反映人們如今在 Google 上實際的跨場景購物方式。這讓零售商在維持與顧客關係的同時，也能協助促成銷售並避免購物車被放棄。

Moving to YouTube. We view YouTube as a flywheel of creators, viewers, and advertisers. First, creators. Every day, more than 20 million videos are being uploaded. We are making creation easier with new generative AI tools in YouTube Studio, plus a conversational editing assistant in Shorts and YouTube create.

接著談 YouTube。我們將 YouTube 視為由創作者、觀眾與廣告主共同驅動的飛輪。第一，創作者。每天有超過 2,000 萬支影片被上傳。我們透過 YouTube Studio 中新的生成式 AI 工具，以及 Shorts 與 YouTube Create 中的對話式剪輯助理，讓創作變得更容易。

Looking at viewership. At Made on YouTube, we introduce short series where creators can structure their shorts into seasons and episodes. Sequential playback makes it seamless for viewers to discover new shorts and keep watching.

再看觀看端。在 Made on YouTube 上，我們推出短篇系列（short series），讓創作者可以把 Shorts 以「季」與「集」的形式來編排。連續播放（sequential playback）讓觀眾能順暢地發掘新的 Shorts 並持續觀看。

The third part is monetization, starting with ads. Our focus remains on delivering value across the entire funnel. In Q3, Living Room accounted for over one-third of total YouTube ads revenues. Small and medium advertisers continue to adopt demand Gen. Driving direct response.

第三部分是變現，先從廣告開始。我們的重點仍是為整個漏斗各階段提供價值。在第三季，客廳場景（Living Room）貢獻了 YouTube 廣告總營收的三分之一以上。中小型廣告主持續採用 Demand Gen，以推動直接回應（direct response）。

On YouTube, Wayfair's demand Gen Campaigns drove a 2.3 times higher return on ad spend compared with traditional video reach campaigns. Our YouTube Creator Partnerships program connects brands with over 3 million creators to drive highly effective campaigns. Brands like Lexus continue to partner with creators like Colin and Samir, driving an 11% lift in brand consideration.

在 YouTube 上，Wayfair 的 Demand Gen 活動相較於傳統的影片觸及活動，帶來高出 2.3 倍的廣告投資報酬率（ROAS）。我們的 YouTube Creator Partnerships 計畫將品牌與超過 300 萬名創作者連結，推動高效的行銷活動。像 Lexus 這樣的品牌持續與 Colin and Samir 等創作者合作，使品牌考慮度提升 11%。

More than 1.3 million channels are enrolled in YouTube Shopping globally. Building on this momentum, we recently announced that Amazon is joining YouTube Shopping. US creators can now tag Amazon products directly in their videos, Shorts and live streams, driving direct conversions.

全球已有超過 130 萬個頻道加入 YouTube Shopping。在這股動能之上，我們近期宣布 Amazon 將加入 YouTube Shopping。美國創作者現在可在其影片、Shorts 與直播中直接標記 Amazon 產品，帶動直接轉換。

Importantly, our subscription ecosystem continues to complement our advertising business. Subscriptions continue to grow faster than ads, led by YouTube Music Premium. YouTube Premium has more than 150 million subscribers. With Premium and Premium Lite, viewers have choice. As more people opt into Premium experiences, it's a win for viewers, creators, and our business. Subscriptions help fund the creator ecosystem and drive the platform forward.

重要的是，我們的訂閱生態系持續與廣告業務相輔相成。訂閱業務持續以快於廣告的速度成長，其中以 YouTube Music Premium 帶動為主。YouTube Premium 的訂閱用戶已超過 1.5 億。透過 Premium 與 Premium Lite，觀眾擁有選擇。隨著更多人選擇 Premium 體驗，這對觀眾、創作者以及我們的業務都是雙贏。訂閱有助於資助創作者生態系並推動平台向前發展。

Finally, strategic partnerships. Businesses are choosing Google because we bring together cloud, AI, ads and consumer platforms to accelerate their transformation. We expanded our partnership with American Express to support the ongoing modernization of their contact centers and collaborate on agentic commerce across Search and Wallet.

最後是策略夥伴關係。企業選擇 Google，是因為我們整合雲端、AI、廣告與消費者平台，加速其轉型。我們擴大與 American Express 的合作，以支援其聯絡中心持續現代化，並在 Search 與 Wallet 上就代理式商務（agentic commerce）展開協作。

And Unity signed an expanded multi-year cloud deal to use our full AI stack to power the next generation game engine. Our commercial priorities are clear, expanding advertiser ROI, systematically advancing our AI monetization efficiency, and scaling our ecosystem partnerships to support the next area of discovery.

此外，Unity 簽署了一項擴大的多年期雲端合作協議，將使用我們完整的 AI 技術堆疊來驅動下一代遊戲引擎。我們的商業優先事項很明確：提升廣告主 ROI、系統性推進 AI 變現效率，以及擴大生態系夥伴合作以支援下一個探索領域。

In closing, I want to thank Googlers Everywhere for their incredible execution and our customers and partners for their continued trust.

最後，我要感謝全球各地的 Googlers 以卓越的執行力完成工作，也感謝我們的客戶與合作夥伴持續的信任。

Thomas, over to you.

Thomas，交給你。

Our strong momentum is driven by our integrated AI portfolio of chips, models, data, security, and agent platforms, all designed to work together. We have strong diversification across our product lines, industries, and customers.

我們強勁的動能來自整合式 AI 產品組合，涵蓋晶片、模型、資料、安全與代理平台，且皆以協同運作為設計目標。我們在產品線、產業與客戶方面具備高度多元化。

Google Cloud now has 18 product lines with more than $1 billion each in annual revenue. And our growth this quarter was driven by strong growth in Google Cloud Platform with strength in both AI and core GCP. TPU system sales were also a meaningful contributor.

Google Cloud 目前有 18 條產品線，每條年營收皆超過 10 億美元。本季成長主要由 Google Cloud Platform 的強勁成長帶動，AI 與核心 GCP 皆表現強勁。TPU 系統銷售亦是重要貢獻來源。

Over the past 12 months, more than 2,600 customers have used at least 100 billion tokens and nearly 75, at least 10 trillion tokens on Gemini Enterprise. Nearly 90% of the Fortune 100 use Gemini Enterprise and customers who use Gemini Enterprise in Q3 2025 more than quadrupled their spend with us year over year.

在過去 12 個月中，超過 2,600 家客戶至少使用了 1,000 億個 token，且將近 75 家客戶在 Gemini Enterprise 上至少使用了 10 兆個 token。《財富》100 大企業中近 90% 使用 Gemini Enterprise，而在 2025 年第三季使用 Gemini Enterprise 的客戶，其對我們的支出年增超過四倍。

One of the important ways in which customers are using Gemini is for knowledge work. Across all our AI services, we have sold more than 100 million seats for knowledge work. Gemini Enterprise now includes the Gemini agent that is deeply integrated across our entire cloud portfolio.

客戶使用 Gemini 的一個重要方式是知識型工作。在我們所有 AI 服務中，已售出超過 1 億個知識工作席次。Gemini Enterprise 現已包含 Gemini 代理（agent），並深度整合於我們整個雲端產品組合之中。

It can research and answer questions, write and run code, and automate process flows with one agent, one API, and a single prompt window. It gives companies strong security and cost controls and the widest choice of frontier models.

它可透過單一代理、單一 API 與單一提示視窗，進行研究並回答問題、撰寫與執行程式碼，以及自動化流程。它為企業提供強大的安全性與成本控管，以及最廣泛的前沿模型選擇。

Customers include BNP Paribas, who is deploying Gemini Enterprise to more than 65,000 employees to streamline workflows while maintaining strict enterprise-grade security and data governance Tata Steel, who deployed more than 300 specialized AI agents in just nine months to enhance manufacturing safety, optimize supply chains, and cut service times for customers by 50%.

客戶包括 BNP Paribas，正將 Gemini Enterprise 部署給超過 65,000 名員工，以在維持嚴格企業級安全與資料治理的同時簡化工作流程；以及 Tata Steel，在短短九個月內部署了超過 300 個專用 AI 代理，以提升製造安全、優化供應鏈，並將客戶服務時間縮短 50%。

Ulta Beauty, who built the Ulta AI shopping assistant, resulting in a 3x higher conversion rate than traditional browsing on their site. Gemini is also driving acceleration for our database, analytics and cybersecurity products.

Ulta Beauty 建置了 Ulta AI 購物助理，使其轉換率較網站上的傳統瀏覽方式高出 3 倍。Gemini 也正在加速我們的資料庫、分析與資安產品。

Gemini can now help users find and visualize information from structured and unstructured data. Gemini uses our knowledge catalog to find the right sources and definitions of information improving accuracy, lowering token cost, and making responses more reliable.

Gemini 現在可協助使用者從結構化與非結構化資料中尋找並視覺化資訊。Gemini 透過我們的知識目錄（knowledge catalog）找出正確的資訊來源與定義，提升準確性、降低 token 成本，並使回應更可靠。

Gemini uses our Lakehouse to access information from multiple systems with zero copy. To respond to a genetic analysis, BigQuery improved query speed over 40% year over year, while reducing query processing cost by over 50% year over year. Customers include Deutsche Telekom, which uses BigQuery, Cloud Storage, and Knowledge Catalog to build a unified data Lakehouse across 40 plus legacy on-premise systems.

Gemini 使用我們的 Lakehouse 以零拷貝（zero copy）方式存取多個系統的資訊。針對一項基因分析回應，BigQuery 的查詢速度年增超過 40%，同時查詢處理成本年減超過 50%。客戶包括 Deutsche Telekom，使用 BigQuery、Cloud Storage 與 Knowledge Catalog，在 40 多個既有的內部部署系統之上建置統一的資料 Lakehouse。

Etsy, which migrated three petabytes of data to unified Lakehouse. By using BigQuery and Iceberg, they now perform critical data joins up to 60% faster. The growing risk of cybersecurity threats is also driving strong demand for our AI-powered security platform.

Etsy 已將 3PB 的資料遷移至統一的 Lakehouse。透過使用 BigQuery 與 Iceberg，他們現在可將關鍵資料 join 的速度提升最多 60%。資安威脅風險日益升高，也正帶動市場對我們 AI 驅動安全平台的強勁需求。

Our Gemini Flash cyber models are frontier models that automate vulnerability discovery across 20 programming languages and patch 2.6 times more reliably than other models. And there's even better performance to come as we roll out Argon and Carbit.

我們的 Gemini Flash 資安模型屬於前沿模型，可在 20 種程式語言中自動化發現弱點，且修補可靠度較其他模型高出 2.6 倍。隨著我們推出 Argon 與 Carbit，未來還會有更好的效能表現。

Our security platform integrates threat intelligence, cyber response prioritization, automated scanning, code remediation and monitoring. Gemini and Google threat intelligence autonomously analyzes millions of external events every day with 98% accuracy. And more than 95% of Wiz customers are adopting AI-powered features and growth continuum.

我們的安全平台整合威脅情報、資安回應優先順序、自动化掃描、程式碼修復與監控。Gemini 與 Google 威脅情報每天可自主分析數百萬個外部事件，準確率達 98%。此外，超過 95% 的 Wiz 客戶正在採用 AI 驅動功能與成長連續體（growth continuum）。

These security tools protect critical infrastructure, including financial institutions like Barnorte and Jack Henry and telecommunication providers such as AT&T. All of this is powered by our leading AI infrastructure spanning our TPU systems, Axion CPUs, the latest Nvidia GPUs, and our Agent Compute Engine, which is specifically designed for AI agents.

這些安全工具可保護關鍵基礎設施，包括 Barnorte、Jack Henry 等金融機構，以及 AT&T 等電信業者。所有這些都由我們領先的 AI 基礎設施所驅動，涵蓋 TPU 系統、Axion CPU、最新的 Nvidia GPU，以及專為 AI 代理設計的 Agent Compute Engine。

TPU systems are specialized systems built for AI combining hardware, compute systems components, and optimized software for training and inference. TPU 8T and 8I offer up to 2.7 times better training performance and 80% better inference price performance, respectively, versus our previous generation of TPUs.

TPU 系統是為 AI 打造的專用系統，結合硬體、運算系統元件，以及針對訓練與推論最佳化的軟體。相較於上一代 TPU，TPU 8T 的訓練效能最高可提升 2.7 倍，而 TPU 8I 的推論性價比最高可提升 80%。

To make it easy for developers to port workloads to TPUs, we recently worked closely with Meta to deliver Torch TPU, which allows users to train, serve, and deploy piles on TPUs using the tools they're already familiar with and rely on.

為了讓開發者更容易將工作負載移植到 TPU，我們近期與 Meta 密切合作推出 Torch TPU，讓使用者可使用既有熟悉且仰賴的工具，在 TPU 上進行訓練、提供服務與部署。

Nearly 100 customers use TPUs today. And we're seeing strong and growing demand from many others. In addition to our work with Anthropic, both directly and via Broadcom and public reference customers like Apple who are scaling pre-training of their foundation models on the latest generation of TPUs, we're also working with new frontier model companies, including Mirendil and Physical Intelligence, capital markets firms including Citadel Securities and Jump Trading, universities like Caltech, and public sector service providers including [Hassava and Cirruskill].

目前已有近 100 家客戶使用 TPU。我們也看到許多其他客戶的需求強勁且持續成長。除了與 Anthropic 的合作（包含直接合作，以及透過 Broadcom），以及像 Apple 這類公開參考客戶正在最新一代 TPU 上擴大其基礎模型的預訓練之外，我們也正與新的前沿模型公司合作，包括 Mirendil 與 Physical Intelligence；資本市場公司包括 Citadel Securities 與 Jump Trading；大學如 Caltech；以及公共部門服務供應商，包括 [Hassava and Cirruskill]。

Given continued strong and accelerating demand, we've recently entered into partnerships to deliver TPU systems with both Crux AI and Oracle with more in the works.

鑑於需求持續強勁且加速成長，我們近期已與 Crux AI 與 Oracle 建立合作夥伴關係以交付 TPU 系統，且還有更多合作正在進行中。

Over to you, Anat.

交給你，Anat。

Thank you. Comments will focus on year-over-year comparisons for the third quarter unless they state otherwise.

謝謝。除非另有說明，以下評論將聚焦於第三季的年對年比較。

I will start with the results at the Alphabet level, and we'll then cover our segment results. I'll end with some commentary in our outlook for the fourth quarter and full year 2026. We had a strong third quarter, delivering our 13th consecutive quarter of double-digit revenue growth.

我將先從 Alphabet 層級的結果開始，接著說明各事業部門的表現。最後我會就第四季與 2026 全年的展望提供一些評論。我們第三季表現強勁，達成連續第 13 個季度的雙位數營收成長。

Consolidated revenues were $127.6 billion, up 25% in both reported and constant currency. Total cost of revenue was $51.4 billion, up 24%. TAC was $16.4 billion, up 10%. Other cost of revenues was $35.1 billion, up 32%, driven by hardware costs primarily related to the sales of depreciation, content acquisition costs, and technical infrastructure operations costs.

合併營收為 1,276 億美元，以報告基礎與固定匯率基礎計算皆年增 25%。營收成本總額為 514 億美元，年增 24%。TAC（流量取得成本）為 164 億美元，年增 10%。其他營收成本為 351 億美元，年增 32%，主要受硬體成本（主要與折舊銷售相關）、內容取得成本以及技術基礎設施營運成本所帶動。

Total operating expenses were up 20% to $35.6 billion. R&D expenses increased by 36%, primarily for investment in AI talent as well as depreciation. Sales and marketing expenses were up 22% due to investments in Gemini and compensation. And G&A expenses decreased 17%, primarily due to lapping of higher legal expenses that occurred in Q3 last year, partially offset by current year legal charges.

營業費用總額年增 20% 至 356 億美元。研發費用增加 36%，主要用於投資 AI 人才以及折舊。銷售與行銷費用年增 22%，係因對 Gemini 的投資及薪酬所致。而一般及管理（G&A）費用下降 17%，主要因去年第三季發生的較高法律費用基期效應，部分被本年度法律費用所抵銷。

Operating income increased 30% to $40.6 billion and operating margin was 31.8%. Other income and expenses was negative $8.2 billion due primarily to unrealized loss from our investment in SpaceX from a decline in its stock price in the third quarter.

營業利益年增 30% 至 406 億美元，營業利益率為 31.8%。其他收益與費用為負 82 億美元，主要因我們對 SpaceX 的投資在第三季因股價下跌而產生未實現損失。

Net income and earnings per share decreased due to the decline in OI&E. We generated strong operating cash flow of $42.3 billion in the third quarter and $179.6 billion for the trailing 12 months. Operating cash flow was negatively impacted by the buildup of inventory for TPU system sales, as well as a $5.2 billion cash payment for a 2018 EC fine following the conclusion of the appeal process.

由於其他收益與費用（OI&E）下降，淨利與每股盈餘下滑。我們在第三季產生強勁的營業現金流 423 億美元，過去 12 個月合計為 1,796 億美元。營業現金流受到為支援 TPU 系統銷售而增加庫存的影響，以及在上訴程序結束後就 2018 年歐盟執委會（EC）罰款支付 52 億美元現金的負面影響。

CapEx was $51.4 billion in the third quarter with a vast majority of the spend in technical infrastructure to support our investments in AI. Approximately 65% of our investments in technical infrastructure in Q3 was related to servers and networking equipment, which generally have useful lives of six years. The remaining 35% was for data centers which have useful lives of 7 to 40 years.

第三季資本支出（CapEx）為 514 億美元，其中絕大部分投入於技術基礎設施，以支援我們在 AI 的投資。第三季技術基礎設施投資約 65% 與伺服器及網路設備相關，其一般使用年限為 6 年；其餘 35% 用於資料中心，使用年限為 7 至 40 年。

We had negative free cash flow of $9.1 billion in the third quarter, driven by our investments in CapEx. Free cash flow in the third quarter was also affected by the aforementioned inventory bill to support TPU system sales and payment of the EC fine.

第三季自由現金流為負 91 億美元，主要由於我們在資本支出上的投資。第三季自由現金流亦受到前述為支援 TPU 系統銷售的庫存支出，以及支付 EC 罰款的影響。

Free cash flow was $19.7 billion for the trailing 12 months. We ended the quarter with $265.1 billion in cash and marketable securities, which includes $90.4 billion of marketable equity securities. Long-term debt was $125.1 billion.

過去 12 個月自由現金流為 197 億美元。本季末我們持有 2,651 億美元的現金及有價證券，其中包含 904 億美元的可交易股權證券。長期負債為 1,251 億美元。

We have also entered into multi-year agreements to secure capacity for technical infrastructure and inventory as well as leases and energy. At the end of the third quarter, we had future purchase commitments of $924 billion and pre-commencement leases and other obligations of $152 billion.

我們亦已簽訂多年期協議，以確保技術基礎設施與庫存的產能，以及租賃與能源供應。第三季末，我們的未來採購承諾為 9,240 億美元，開工前租賃及其他義務為 1,520 億美元。

These dollar amounts reflect the maximum contractual obligations. However, a number of these agreements have variable pricing and volume flexibility, which could result in lower spend. And as mentioned in our press release, our Board of Directors declared a quarterly cash dividend on our common stock of $0.22 per share, which is payable in December.

上述金額反映最大合約義務。然而，其中若干協議具有可變動定價與數量彈性，可能導致實際支出較低。此外，如同我們在新聞稿中提到的，董事會宣告普通股每股 0.22 美元的季度現金股利，將於 12 月支付。

Turning to segment results, Google Services revenues increased 11% to $96.3 billion, reflecting strong growth in search and subscriptions. Total advertising revenues were up 11% with growth across all major verticals. As we stated in our outlook last quarter, FX was a slight headwind on year-over-year growth rates and had a more significant impact on sequential basis.

接著看部門表現，Google 服務營收年增 11% 至 963 億美元，反映搜尋與訂閱的強勁成長。廣告營收總額年增 11%，所有主要垂直產業皆有成長。如同我們上季在展望中所述，外匯（FX）對年對年成長率略為不利，對季對季的影響則更為顯著。

Google Search and other advertising revenues increased by 12% to $63.3 billion with retail driving the largest contribution followed by finance. YouTube advertising revenues increased 9% to $11.2 billion with brands driving the largest contribution and notable strength in the living room.

Google 搜尋及其他廣告營收年增 12% 至 633 億美元，其中零售貢獻最大，其次為金融。YouTube 廣告營收年增 9% 至 112 億美元，品牌廣告貢獻最大，且在客廳場景（living room）表現尤為強勁。

Network advertising revenues of $7.5 billion were up 1%. Subscription platforms and devices revenues increased 11% this quarter to $14.3 billion, driven by strong growth in YouTube subscriptions, particularly in YouTube Music and Premium, and strong demand for AI plans in Google One.

聯播網（Network）廣告營收為 75 億美元，年增 1%。訂閱、平台與裝置營收本季年增 11% 至 143 億美元，主要由 YouTube 訂閱（尤其是 YouTube Music 與 Premium）的強勁成長所帶動，且 Google One 的 AI 方案需求強勁。

Google services operating income increased 17% to $39.2 billion, and operating margin was 40.7%. The Google Cloud segment delivered record growth in the third quarter. Cloud revenues were up 103% to $30.7 billion, driven by strong growth in GCP with strength in AI, including both AI infrastructure and AI solutions and core GCP.

Google 服務營業利益年增 17% 至 392 億美元，營業利益率為 40.7%。Google Cloud 部門在第三季創下歷史新高的成長。雲端營收年增 103% 至 307 億美元，主要由 GCP 的強勁成長所帶動，AI（包含 AI 基礎設施與 AI 解決方案）以及核心 GCP 均表現強勁。

TPU system sales were also a meaningful contributor. GCP accounted for the vast majority of the revenue generated in Q3. The cloud revenue growth rate, excluding the impact of TPU systems was again incredibly strong.

TPU 系統銷售亦是重要的貢獻來源。第三季所產生的營收絕大部分來自 GCP。若排除 TPU 系統的影響，雲端營收成長率仍然非常強勁。

Cloud operating income was $11.1 billion, more than tripling year-over-year. Operating margin increased from 23.7% in the third quarter last year to 36.2%. The third-party bridging capacity, which we discussed on our prior call, came online only at the very end of Q3 and therefore had negative impact to our margin in the quarter.

雲端營業利益為 111 億美元，年對年成長超過三倍。營業利益率由去年第三季的 23.7% 提升至 36.2%。我們在前次電話會議中提到的第三方橋接產能（bridging capacity）直到第三季末才上線，因此在本季對我們的利潤率造成負面影響。

Google Cloud's backlog was $561 billion in the third quarter, with the majority of the backlog related to typical GCP contracts. We expect to recognize just over 55% of the total backlog as revenues over the next 24 months.

第三季 Google Cloud 的待履約合約（backlog）為 5,610 億美元，其中大部分與典型的 GCP 合約相關。我們預期在未來 24 個月內，將把總待履約合約中略高於 55% 認列為營收。

In other bets, revenues were $372 million and the operating loss was $2 billion as we continue to expand Waymo's business and invest in key other bets. In an Alphabet level activities, the operating loss was $7.7 billion, driven by shared AI R&D expenses.

在其他投資（Other Bets）方面，營收為 3.72 億美元，營業虧損為 20 億美元，因我們持續擴大 Waymo 的業務並投資於其他關鍵項目。在 Alphabet 層級活動方面，營業虧損為 77 億美元，主要由共同承擔的 AI 研發費用所致。

I'll now turn to our outlook for the remainder of 2026. First, in terms of revenues. The current spot rates, we would expect a slightly larger FX headwinds to our consolidated revenues in Q4 relative to the prior quarter. This impact will be seen primarily in Search and YouTube ads.

接下來我將談談 2026 年剩餘期間的展望。首先，就營收而言。以目前即期匯率來看，我們預期第四季合併營收所受的外匯逆風將較前一季略大。此影響將主要反映在搜尋與 YouTube 廣告。

In Google Services, we will continue to lap the Search performance acceleration that began in the third quarter last year, which will pressure revenue growth in Q4. In Google Cloud, we expect TPU system sales to continue to scale in Q4, followed by a meaningful revenue ramp in 2027.

在 Google 服務方面，我們將持續面臨去年第三季開始的搜尋表現加速所形成的基期效應，這將對第四季營收成長造成壓力。在 Google Cloud 方面，我們預期 TPU 系統銷售將在第四季持續擴大規模，並在 2027 年出現顯著的營收爬坡。

With respect to Cloud margins, given the supply-constrained environment, we started utilizing third-party compute capacity at the end of Q3 and will ramp usage in Q4. We expect margin headwinds in the near term as we utilize this capacity.

就雲端利潤率而言，鑑於供給受限的環境，我們在第三季末開始使用第三方運算產能，並將在第四季提高使用量。隨著我們使用該產能，我們預期短期內利潤率將面臨逆風。

Moving to investments, our full year 2026 CapEx guidance range of $195 billion to $205 billion remain unchanged. We continue to expect our CapEx to increase significantly in 2027 and will provide further details on our Q4 call.

談到投資，我們 2026 全年資本支出指引區間 1,950 億至 2,050 億美元維持不變。我們仍預期 2027 年資本支出將大幅增加，並將在第四季電話會議提供更多細節。

In terms of expenses, the significant increase in our investment in technical infrastructure will continue to weigh on the P&L in the form of higher depreciation expenses and related data centers operation costs such as energy.

就費用而言，我們在技術基礎設施上的投資大幅增加，將持續以較高折舊費用及相關資料中心營運成本（如能源）等形式，對損益表（P&L）造成壓力。

Before we move to Q&A, let me say a few words on how we think about return on invested capital and free cash flow. We're in a unique position to be able to monetize our AI investments across many products that reach billions of users and enterprises.

在進入問答之前，我想先談幾句我們如何看待投入資本報酬率與自由現金流。我們處於獨特的位置，能夠在多項觸及數十億使用者與企業的產品上，將我們的 AI 投資變現。

Our AI investments in Search are driving our highest query growth in half a decade. Google Cloud has generated tens of billions of dollars in AI revenues year-to-date and Google One is our fastest-growing subscription products driven by demand for AI plans.

我們在搜尋領域的 AI 投資正推動過去半個十年來最高的查詢成長。Google Cloud 今年迄今已創造數百億美元的 AI 收入，而在 AI 方案需求帶動下，Google One 是我們成長最快的訂閱產品。

Throughout this investment cycle, we expect to generate an attractive ROIC that remains meaningfully above our cost of capital and anticipate generating strong cash flow from operations. Our sizable investments in technical infrastructure and inventory build of TPU systems as we scale that business will put pressure on free cash flows.

在這一輪投資週期中，我們預期將產生具吸引力的投資資本報酬率（ROIC），且將顯著高於我們的資本成本，並預期營運活動將帶來強勁現金流。隨著我們擴大該業務規模，對技術基礎設施的大量投資以及 TPU 系統的庫存建置，將對自由現金流造成壓力。

Q3 was a strong quarter highlighted by our innovation and disciplined execution across all areas of the business, including the launch of Gemini 4. I want to take this opportunity to thank our employees for their contributions to our performance.

第三季表現強勁，亮點在於我們在業務各領域的創新與紀律執行，包括推出 Gemini 4。我也想藉此機會感謝員工對我們業績表現的貢獻。

We will now open the call for Q&A.

我們現在開始 Q&A 問答環節。

(Operator Instructions)

（接線員指示）

Doug Anmuth, JPMorgan.

Doug Anmuth，摩根大通。

Thanks for taking the questions. One for Sundar and one for Philip. Sundar, on a recent podcast, you said that 2027 could be the golden age of productivity for Google because of the new class of models and how they'll integrate with your other products. I was hoping you could elaborate on that and just share more how you think about the opportunity set there.

謝謝讓我提問。我有一題問 Sundar，一題問 Philip。Sundar，你在最近的一個 Podcast 中提到，因為新一代模型以及它們將如何與其他產品整合，2027 年可能會是 Google 生產力的黃金時代。我希望你能進一步闡述，並分享你如何看待其中的機會版圖。

And then, Philip, just on query growth, the fastest in five years, but obviously some deceleration in search growth. Can you just help us understand the disconnect there? Is it simply that you haven't been showing as many ads basically in AI overviews and AI mode, and that's now becoming a bigger percentage of the overall queries? Thanks.

另外，Philip，關於查詢成長，這是五年來最快，但顯然搜尋營收成長有些放緩。你能幫我們理解兩者之間的落差嗎？是否只是因為在 AI 摘要（AI overviews）與 AI 模式中，你們基本上沒有展示那麼多廣告，而這些現在正占整體查詢更高的比例？謝謝。

Thanks, Doug. We definitely started seeing a step change, I would say, with 3.8 Flash internally as our models got better at long horizon tasks in the coding, et cetera. But definitely seeing a huge point of inflection once we have started deploying Argon internally.

謝謝，Doug。我們確實開始看到一個階躍式的變化；我會說，當 3.8 Flash 在內部使用時，隨著模型在程式撰寫等長期跨度任務上的能力變得更好。但更明顯的拐點是在我們開始於內部部署 Argon 之後。

It is one of the most meaningful internal productivity levers we have seen on software engineering directly tracking high-quality code that's getting checked in, the rate at which it's getting checked in, reducing the number of rollbacks. But more importantly, people are tackling massive problems as well as we are beginning to see innovation on the product and the consumer experience side from deploying those models.

這是我們在軟體工程方面所見過最具意義的內部生產力槓桿之一：直接追蹤被提交（check-in）的高品質程式碼、提交速度，並降低回滾（rollback）的次數。但更重要的是，隨著我們部署這些模型，我們也看到人們開始處理更龐大的問題，同時在產品與消費者體驗端也開始出現創新。

But all of this happened only about four weeks ago. We are now deploying carbon internally and carbon is a meaningful step up from Argon. So as we roll it out and make it all widely available to Googlers with more compute, I expect to that's what I was referring to and I'm even more excited since I said that in the podcast and what I would be was a golden year ahead in terms of us being able to be more productive and so excited for the innovation ahead.

不過，這一切其實大約是四週前才發生的。我們現在正在內部部署 Carbon，而 Carbon 相較於 Argon 是一個顯著的升級。因此，當我們逐步推廣、並在更多算力支援下讓所有 Googler 都能廣泛使用時，我預期——這就是我在 Podcast 裡所指的；而且自從我在 Podcast 說完之後我更興奮了——未來將會是一個黃金年，因為我們能更有效率，也非常期待接下來的創新。

Phillip?

Phillip？

Doug, on the second part of your question, in Q3, search grew 12% year-over-year to $63 billion. And when you look at the change in growth rate from Q2 to Q3, there were a few drivers here. So first, there was FX. It was the single largest driver of search deceleration relative to Q2.

Doug，針對你問題的第二部分，第三季搜尋營收年增 12% 至 630 億美元。而當你看第二季到第三季成長率的變化時，背後有幾個驅動因素。首先是匯率（FX），相較於第二季，這是造成搜尋成長放緩的最大單一因素。

Second, there were vertical-specific dynamics. In retail, some of our APAC advertisers were impacted by the EU eliminating its de minimis exemption. In media and entertainment spend normalized after an unseasonably strong Q2 driven by the FIFA World Cup, which had more matches and therefore more advertising budgets allocated in June than actually in July.

第二，是特定垂直領域的動態。在零售方面，我們部分亞太地區（APAC）的廣告客戶受到歐盟取消小額免稅（de minimis）豁免的影響。在媒體與娛樂方面，支出在第二季因 FIFA 世界盃帶動而異常強勁後回歸常態；該屆賽事場次更多，因此 6 月分配的廣告預算反而比 7 月更多。

It's also to the next part of your question, it's worth noting that the strong growth in AI mode queries had an impact because of the low monetization we were seeing in AI mode. To be clear, we feel very good about our trajectory here. Our ad experiments have shown promising results, and we're now scaling ads within both AI Overuse and AI Mobile to more users and more queries, including, as I mentioned, international markets and multi-turnouts in the US.

另外也呼應你問題的下一部分，值得注意的是，AI 模式查詢的強勁成長確實帶來影響，因為我們在 AI 模式中看到的變現程度較低。但要說清楚，我們對這裡的發展軌跡感到非常樂觀。我們的廣告實驗已展現令人鼓舞的結果，而我們現在正把 AI Overuse 與 AI Mobile 中的廣告擴大推廣給更多使用者與更多查詢；包括我提到的國際市場，以及美國的多輪互動（multi-turnouts）。

Thank you.

謝謝。

Brian Nowak, Morgan Stanley.

Brian Nowak，摩根士丹利。

Thanks for taking my questions. I have two. Maybe let me start with Agentic. You rolled out Spark to some of the users throughout the year. I know, Sundar, you've talked a lot about shipping more Agentic products to the broader base of users. Can you just help us understand a little bit some of the key hurdles you have to overcome technologically to ship out Agentic products to the full base of users? Is it model efficiency? Is it capacity? Is it the technology isn't ready? Like what do we have to do technologically to do that.

謝謝讓我提問。我有兩個問題。我先從 Agentic 談起。你們今年把 Spark 推出給部分使用者。Sundar，我知道你談了很多要把更多 Agentic 產品推向更廣泛的使用者群。你能否幫我們理解一下，從技術面要把 Agentic 產品推向全體使用者，主要需要克服哪些關鍵障礙？是模型效率嗎？是容量嗎？還是技術尚未成熟？我們在技術上需要做到什麼才能實現？

And then the second one, just on to Doug's question on search, Anat, can you help us better understand the search performance comparison factor that you talked about impacting search, and how do we think about that as we kind of go into 4Q, 1Q, just kind of looking at the compares? Thanks.

第二個問題，延續 Doug 對搜尋的提問，Anat，你能否幫我們更好地理解你提到的、影響搜尋的「搜尋表現比較因素」，以及當我們進入第四季、第一季、看同比基期（compares）時，應該如何看待它？謝謝。

Thanks. Thanks, Brian. First of all, on agents, what an exciting and dynamic area. I think when I look a our new models and we see the capabilities of the kind of complex tasks they can take in an automated way and execute with very high-quality. It's clear that the agent capabilities we've all envisioned are going to be real and so I couldn't be more excited to bring the next generation agent experiences to our users.

謝謝。謝謝，Brian。首先談代理（agents），這是一個令人振奮且變化快速的領域。當我看我們的新模型，並看到它們能以自動化方式承接並以非常高品質執行各類複雜任務的能力時，很明顯，我們一直想像的代理能力將會成真，因此我對把下一代代理體驗帶給使用者感到無比興奮。

To your question, I would say a few things. I think the Gemini for model series has been important. We are working on distant versions of our carbon models so that we can deploy it across a billion user scale efficiently, both from a compute, from a latency standpoint and from a quality standpoint. So that's been an important area of investment. Definitely making sure you need more than a capable model at scale, you need a robust harness, you need a highly secure runtime, and this is where GCP comes in, being able to give each and every one of our users a secure runtime of availability in the cloud, the right ecosystem of skills and connectors.

針對你的問題，我會說幾點。我認為 Gemini 4 模型系列一直很重要。我們正在開發 Carbon 模型的精簡版本，以便能在十億使用者規模下高效率部署，無論是在算力、延遲或品質層面。因此這一直是重要的投資領域。此外，在大規模情境下，你需要的不只是能力強的模型；你還需要健全的支撐框架（harness）、高度安全的執行環境（runtime）。這也是 GCP 發揮作用之處：能在雲端為我們每一位使用者提供安全且可用的執行環境，以及合適的技能與連接器生態系。

So all that is the foundation. We have been hard at work, and we decided to roll it out to Enterprise first, which we announced with our Gemini Agent for Enterprise essentially because they have the highest bar for security governance and controls and it's been well received and we are in the process of expanding that.

因此，這些都是基礎。我們一直在全力投入，也決定先向企業客戶推出；我們已宣布推出 Gemini Agent for Enterprise，主要是因為企業對安全治理與控管的門檻最高。這項產品反應良好，我們也正在擴大推廣。

Later this year, we will bring that same experience to consumers, starting with the 1 billion users in our Gemini app. We already see people in the app moving beyond chat to what task and agent-like experiences and so the trust users have in our products, the fact that all the context is right there, I think sets us up well, but we are really focused as a company on delivering that unified experience across our products and you should expect to see a lot of progress later this year.

今年稍晚，我們會把同樣的體驗帶給消費者，首先從我們 Gemini 應用程式的 10 億使用者開始。我們已經看到使用者在 App 中不再只停留在聊天，而是走向任務型與類代理（agent-like）的體驗；再加上使用者對我們產品的信任、所有情境脈絡都在那裡，我認為這讓我們具備良好條件。不過，我們公司目前非常聚焦於在各產品之間交付一致的整合體驗，你們應該會在今年稍晚看到大量進展。

Thanks. And on the second question with regards to my comments, I believe you're referring to the outlook with regards to Search. A couple of dynamics that I've mentioned, one is on sequential related to impact of FX headwinds, which we saw this quarter as well versus the previous quarter. So where we are today with current spot rates, we expect that this will be somewhat headwind to Q4 result.

謝謝。關於第二個問題、針對我剛才的評論，我相信你指的是搜尋（Search）的展望。我提到的幾個動態因素之一，是與外匯逆風影響相關的季度序列變化（sequential），我們本季相較上一季也看到了。因此，以目前的即期匯率水準來看，我們預期這會對第四季（Q4）業績形成一定程度的逆風。

And when we look at the year-over-year comparisons. This quarter, we started lapping the acceleration we saw in Search growth in 2025 at the back end of 2025. We left about 3 points of growth versus Q3. And then we expect to see that continuing in Q4, a couple of points of lapping that growth acceleration that we saw at the end of 2025.

再看年對年（year-over-year）的比較。本季開始進入與 2025 年下半年、也就是 2025 年底我們在搜尋成長加速的那段期間的比較基期（lapping）。相較第三季（Q3），我們少了大約 3 個百分點的成長。

Eric Sheridan, Goldman Sachs.

Eric Sheridan，高盛（Goldman Sachs）。

Thanks so much for taking the question. Based on some of the comments you've made recently, Sundar, on compute constraints, can you refresh for us your key priorities for the application of compute as it becomes available across the organization?

非常感謝讓我提問。Sundar，基於你最近對算力（compute）限制的一些評論，你能否為我們更新一下：隨著算力在組織內逐步可用，你在算力應用上的關鍵優先順序是什麼？

And the second part, maybe for you or Thomas, would be within Google Cloud specifically, how did your view continue to evolve with respect to the application of compute to cloud customer needs relative to external TPU? Thanks so much.

第二部分，可能請你或 Thomas 回答：在 Google Cloud 方面，你們對於將算力用於雲端客戶需求、相對於外部 TPU 的看法，近期是如何持續演進的？非常感謝。

Thanks, Eric. I'll start and then Thomas will answer the second-half. Definitely, computer allocation is an area where we are spending a lot of time on because of the forward planning that's needed. I would start with saying we feel fortunate to have an extraordinary set of opportunities ahead of us.

謝謝你，Eric。我先回答，然後 Thomas 會回答後半段。確實，算力配置（compute allocation）是我們投入大量時間的領域，因為需要做前瞻性的規劃。我先說，我們覺得很幸運，因為眼前有一組非常卓越的機會。

We are operating a scale consumer business that reaches billions of people, a hyperscale cloud business with a lot of momentum and of course, a range of other highly important areas across Google and Alphabet. We always take a long-range planning and so each of our business areas summit long-range plans with a very disciplined ROIC framework, and so broadly think of all of this as a long-range planning process, not a short-term allocation exercise.

我們營運的是一個觸及數十億人的大規模消費者業務、一個動能強勁的超大規模（hyperscale）雲端業務，當然還有 Google 與 Alphabet 旗下一系列其他非常重要的領域。我們一向採取長期規劃，因此各業務領域都會在非常嚴謹的 ROIC（投入資本報酬率）框架下提交長期計畫；所以整體而言，這是一個長期規劃流程，而不是短期的配置練習。

Obviously our AI models underlie everything we do from Search and YouTube to cloud to Waymo and so on. So it's a foundation and we view it as a highly leveraged approach. The full stack approach drives a lot of efficiency as well as gives us a lot of optionality across a wide variety of opportunities.

很明顯，我們的 AI 模型是我們從搜尋與 YouTube 到雲端、Waymo 等各項業務的底層支柱。因此它是基礎，我們也把它視為高度槓桿化的方法。全棧（full stack）方法不僅帶來大量效率，也讓我們在各式各樣的機會上擁有很大的選擇權（optionality）。

So obviously that is the foundation and we plan for that for the long-term too. And about that, as I said, we really look at our businesses and really take a disciplined long-range planning approach. To us, obviously, we are seeing very robust demand signals across a wide variety of areas. And so we view it as an extraordinary set of opportunities and we are investing for that moment.

所以很明顯，這是基礎，我們也以長期角度來規劃。而且如我所說，我們確實是以嚴謹的長期規劃方式來看待各項業務。對我們而言，我們在非常多元的領域都看到非常強勁的需求訊號。因此我們把它視為一組非凡的機會，並為此時刻進行投資。

Thomas?

Thomas？

Eric, on the demand in cloud, we are always managing our portfolio to ensure we have a breadth of customers and we have different types of customers from digital natives, AI labs, traditional enterprises, financial and capital markets customers. So it's a broad set of customers.

Eric，談到雲端需求，我們一直在管理我們的產品組合（portfolio），以確保客戶結構足夠多元；我們的客戶類型包含數位原生（digital natives）、AI 實驗室、傳統企業、金融與資本市場客戶。所以客戶群非常廣泛。

Many of them have either started with using TPU systems, for example, in our cloud. We have over 100 significant TPU customers who start in our cloud, they learn to use TPUs and remember TPU system is a combination of hardware, system software, compilers, the whole thing assembled as a system, including optical networking things.

其中許多客戶一開始就是在我們的雲端使用 TPU 系統。我們有超過 100 家重要的 TPU 客戶從我們雲端起步，他們學習如何使用 TPU；而且請記得，TPU 系統是硬體、系統軟體、編譯器等的組合，是把整套東西整合成一個系統，還包含光學網路等元件。

We've taken a very deliberate approach on how we're bringing TPUs to customers. The first one is to build the right silicon. So we have AT for training, AI for inference. They both offer substantial price performance and performance improvements relative to prior generations.

我們在把 TPU 帶給客戶這件事上採取了非常審慎的做法。第一是打造正確的晶片（silicon）。我們有用於訓練（training）的 AT，以及用於推論（inference）的 AI。相較前一代，它們在性價比（price performance）與效能上都有顯著提升。

Second, we've optimized the software. The optimization software is not just to make the systems run better and to give you great performance, but also to make portability seamless. Recently, we introduced, for example, PyTorch, which opens up a much larger market for us by optimizing PyTorch on the TPU.

第二，我們優化了軟體。這些軟體優化不只是讓系統跑得更好、提供很棒的效能，也讓可攜性（portability）能夠無縫銜接。例如我們最近導入了 PyTorch，透過在 TPU 上優化 PyTorch，為我們打開了更大的市場。

Third, we are broadening distribution. So in the past, you could use TPUs in our cloud. Now we make for customers who want TPU systems in their data centers, we're making TPU systems available in their data centers. We're also broadening availability to other providers.

第三，我們正在擴大分發（distribution）。過去你只能在我們的雲端使用 TPU。現在，對於希望在自家資料中心部署 TPU 系統的客戶，我們也讓 TPU 系統能在他們的資料中心使用。我們也在擴大對其他供應商的可用性。

For example, we announced a partnership with Oracle and we've launched, along with Blackstone, an agreement for Crux AI, a joint venture to offer customers TPU systems. We are seeing extraordinary demand and strong growth from both AI labs, capital markets, financial service institutions and enterprises. And if you imagine these 100 plus customers growing to many more and also scaling their use into TPU systems, we see strong demand over the future.

例如，我們宣布了與 Oracle 的合作夥伴關係；並且與 Blackstone 一起，為 Crux AI 推出一項協議——這是一家合資企業，用來向客戶提供 TPU 系統。我們看到來自 AI 實驗室、資本市場、金融服務機構與企業客戶的非凡需求與強勁成長。如果你想像這 100 多家客戶成長到更多、並且把使用規模擴大到 TPU 系統，我們看到未來仍有強勁需求。

Ross Sandler, Barclays.

Ross Sandler，巴克萊（Barclays）。

Thanks. I wanted to stick with the TPU resale business topic that you just mentioned. So in that last quarter, you kind of said that the revenue will be recognized on system sales, I believe. So if a gigawatt of TPU full system costs around $35 billion for the compute and all the kind of components around it, what percent of that are we actually recognizing as revenue? And what kind of flow through does that have in terms of margin profile relative to where the Google Cloud segment currently sits?

謝謝。我想延續你剛提到的 TPU 轉售（resale）業務話題。所以在上一季，你們有提到我理解是收入會在系統銷售時認列。如果一個 GW（gigawatt）的 TPU 完整系統，包含算力與周邊各種元件，成本大約是 350 億美元，那其中有多少百分比會被我們實際認列為營收？而這對毛利率結構的影響（flow through）會是什麼樣子，相較於 Google Cloud 目前所在的區間？

And then the second part of the question would be, as you help companies like Anthropic bring on more of their own kind of first-party infrastructure, what does that mean for the CapEx intensity of Alphabet as you start to kind of hand off some of the future CapEx to AI labs and may not require as much CapEx intensity on your own, in your own data centers. Thoughts on that would be great. Thank you.

然後問題的第二部分是：當你們協助像 Anthropic 這樣的公司導入更多自有（first-party）基礎設施時，隨著你們把部分未來的資本支出（CapEx）交由 AI 實驗室承擔、你們自家資料中心可能不需要那麼高的資本支出強度（CapEx intensity），這對 Alphabet 的資本支出強度意味著什麼？很希望聽聽你們的看法。謝謝。

Yeah, let me combine these two questions. First, obviously, we don't provide specific financials at the product level. But let me walk through how to think about the financials associated with the TPU system sales. And obviously, just as a reminder, and Thomas can elaborate on that as well, a TPU system is not just a chip, but it's a specialized system that combines multiple things, including hardware, compute, storage, networking. So we are selling much more than just a chip, software, training, inference.

好的，我把這兩個問題一起回答。首先，很明顯我們不會在產品層級提供特定財務數字。但我可以說明一下，該如何理解與 TPU 系統銷售相關的財務。同時也提醒一下——Thomas 也可以補充——TPU 系統不只是晶片，而是一個結合多項要素的專用系統，包括硬體、算力、儲存與網路。所以我們賣的不只是晶片，還包含軟體、訓練與推論。

There's a significant added value that's being provided to the customer that's purchasing a TPU system. And the way this works, since it linked to your second portion of the question, when we sign an agreement with a customer to sell TPU systems, once we sign that agreement, you will first see that showing up in our backlog and then we start building inventory.

對於購買 TPU 系統的客戶，我們提供了顯著的附加價值。而這個運作方式也連結到你問題的第二部分：當我們與客戶簽署銷售 TPU 系統的協議後，一旦簽約完成，你會先看到它反映在我們的訂單積壓（backlog）中，接著我們就開始建立庫存（inventory）。

So think about this from a CapEx versus cash perspective. We build the inventory ahead of getting the revenue stream in. So I think I've alluded to this on the previous call, you see inventory balances go up where previously inventory balances primarily were some of our hardware products. Now those TPU systems that are not yet sold will be showing up, so you see cash going out and we're building up the inventory base.

所以，請從資本支出（CapEx）相對於現金的角度來思考這件事。我們會在收入流進來之前，先把庫存建立起來。所以我想我在上一通電話裡也提到過，你會看到庫存餘額上升；過去庫存餘額主要來自我們的一些硬體產品。現在那些尚未售出的 TPU 系統也會反映在庫存裡，因此你會看到現金流出，而我們正在擴大庫存基礎。

And then as we start delivering and billing customers for these CPU systems, then you see cash coming in. So it's not the traditional kind of CapEx as you're thinking about this, and I think that's the essence of your question, but rather a different cycle of cash flow generation.

接著，當我們開始交付並向客戶開立這些 CPU 系統的帳單時，你就會看到現金流入。因此，這並不是你所想的那種傳統 CapEx，我想這就是你問題的核心；而是另一種現金流生成的週期。

Now, given that we are in the process of building up and ramping up that business, you see that cash needs in our cash flow statements and you see those inventory build. To date, we've had some system sales recorded as revenue, but expect more of it to come in through 2027.

現在，鑑於我們正處於建立並加速擴張該業務的過程中，你會在現金流量表中看到這些現金需求，也會看到庫存的累積。截至目前，我們已將部分系統銷售記錄為營收，但預期更多會在 2027 年之前陸續進來。

Justin Post, Bank of America.

Justin Post，美國銀行。

Thank you. I have some questions on the Agentic transition. You did launch the Universal Shopping Cart. Clearly, people are liking and using Agentic use cases. As more of that happens, how do you plan to manage the search transition to Agentic use? Do you think that could be a headwind?

謝謝。我有一些關於 Agentic 轉型的問題。你們確實推出了 Universal Shopping Cart。很明顯，人們喜歡並正在使用 Agentic 的使用案例。隨著這種情況愈來愈多，你們打算如何管理從搜尋轉向 Agentic 使用的過渡？你認為那可能會成為逆風嗎？

And second, do you think Agentic could actually drive more monetization for your traffic over the long-term? And then I'll go again at TPU margins. We're wondering, do you think is a really good business for you and maybe why? If you can't give us the margins, why you think this is a good way to use all this capacity? Thank you.

第二，你認為從長期來看，Agentic 是否反而能為你們的流量帶來更多變現？然後我還要再問一次 TPU 毛利率。我們想知道，你是否認為這對你們而言是一門非常好的生意，以及原因是什麼？如果你不能提供毛利率數字，那你為什麼認為用這些產能來做這件事是個好方式？謝謝。

We're in the early days of Agentic, but maybe directly to this part of your question, but we view Agentic experiences as clearly additive. Our focus is building trusted, secure, superscale experiences that users love. Frankly on Agenetic commerce foundational logs like you mentioned universal commerce, the card, the agent payments protocols and so on.

我們仍處於 Agentic 的早期階段，但直接回應你問題的這一部分：我們認為 Agentic 體驗顯然是加成的。我們的重點是打造使用者喜愛、可信賴、安全、可超大規模擴展的體驗。坦白說，在 Agentic 商務的基礎層面，例如你提到的通用商務、購物車、代理支付協議等等。

When users shop, that's important. There's a part of the journey they don't like, the grunt work, which is frankly fantastic to automate with AI. But we also want to enhance, not just automate the way the discovery experience that people frankly love in so many cases.

當使用者購物時，這很重要。旅程中有一部分他們不喜歡——那些繁瑣的苦工——而這正是用 AI 來自動化的絕佳場景。但我們也希望強化，而不只是自動化，人們在許多情境下其實很喜歡的探索（discovery）體驗。

Users like choice and they want to consider things way beyond price like brand loyalty, customer service and so on. So we're really architecting these identic workflows in a way that we can intentionally design the time and the space for users to explore those choices. So simple example, an agent could respond, here's the insurance product we found and here's another strong option, one of which is a higher premium, maybe a lower deductible and they're currently offering a very special discount.

使用者喜歡有選擇，他們想考量的遠不只價格，還包括品牌忠誠度、客戶服務等等。因此，我們正在以一種方式來架構這些 Agentic 工作流程，使我們能有意識地為使用者設計時間與空間，讓他們去探索這些選擇。舉個簡單例子，代理可以回應：這是我們找到的保險產品，這裡還有另一個很強的選項；其中一個保費較高、但可能自付額較低，而且目前提供非常特別的折扣。

So, given that there's almost never only one perfect option for a decision, we're actually confident that we can actually provide really interesting, compelling advised models for agentic use cases. And obviously, you can imagine going beyond that with subscriptions and other payment methods. But I'm really actually excited about the advertising opportunities in an agentic world.

所以，鑑於做決策時幾乎不會只有一個完美選項，我們其實很有信心能為 Agentic 使用案例提供非常有趣、具吸引力的建議式模型。而且很顯然，你也可以想像進一步延伸到訂閱制與其他支付方式。但我其實對於在 Agentic 世界中的廣告機會感到非常興奮。

Overall, I would add and say, look, I would view it as expansion opportunity at any time, just like a few years ago, being able to incorporate AI chat use cases in Search allow people to ask more and do more in Search, which is what is playing out in our usage, I think agents would again be an opportunity to expand the types of problems we solve for our users.

整體而言，我會補充說：你看，我會把它視為任何時候的擴張機會；就像幾年前一樣，把 AI 聊天使用案例納入搜尋，讓人們能在搜尋中問更多、做更多——這正是我們在使用量上看到的發展。我認為代理同樣會是一次機會，去擴大我們為使用者解決的問題類型。

I think people always underestimate all the components you need to deliver these complex experiences for users. And Search on the surface looks simple, but there's extraordinary deep levels of investment and things we have built over time, like knowledge graph being an example and our understanding of the world, et cetera. So I think I'm excited about agents and I would view it as a new frontier, which brings in a new set of use cases to Search and so we are excited to work through that.

我認為人們總是低估了要為使用者交付這些複雜體驗所需的所有組成要素。搜尋表面上看起來很簡單，但背後有極其深厚的投資層次，以及我們長期建立的能力，例如知識圖譜（knowledge graph）以及我們對世界的理解等等。所以我對代理感到興奮，並把它視為一個新的前沿，它會為搜尋帶來一組新的使用案例，因此我們也很期待把這件事推進下去。

On TPU systems and margins, we don't break out margins for individual products or infrastructure components. However, we see several benefits from designing and manufacturing TPU systems. So first of all, to remind you the TPU system is not just a chip, but it's a complete system.

關於 TPU 系統與毛利率，我們不會拆分單一產品或基礎設施元件的毛利率。不過，我們從設計與製造 TPU 系統中看到了幾項好處。首先提醒一下，TPU 系統不只是晶片，而是一個完整的系統。

Given the price performance benefits of TPU systems relative to competitors, growing demand for accelerated compute and the scarcity of accelerated compute, we're very focused on continuing to improve margins as we scale the business.

鑑於 TPU 系統相較競品在性價比（price performance）上的優勢、加速運算需求的成長，以及加速運算的稀缺性，我們非常專注於在擴大業務規模的同時，持續改善毛利率。

Two additional points. First of all, when Anat mentioned that we procure components to build inventory, we only do that have signed contracts. So all of the inventory that for TPU systems are associated with committed contracts. Additionally, by putting TPU systems in third-party facilities and in customer data centers, it allows us, one, to broaden our addressable markets.

再補充兩點。第一，當 Anat 提到我們採購元件來建立庫存時，我們只會在已簽署合約的情況下才這麼做。因此，TPU 系統的所有庫存都對應到已承諾的合約。另外，把 TPU 系統部署在第三方設施以及客戶資料中心，讓我們第一，可以擴大可服務市場（addressable markets）。

For some industries, they have data in their premise that they cannot move to a cloud, and it allows us to bring accelerated compute to where the data sets. Additionally, it also offsets the need for us to have pre-built facilities to host these TPU systems.

對某些產業而言，他們的資料在自有機房（on-premise），無法搬到雲端；這讓我們能把加速運算帶到資料所在之處。此外，這也抵消了我們必須預先建置設施來託管這些 TPU 系統的需求。

Mark Shmulik, Alliance Bernstein.

Mark Shmulik，Alliance Bernstein。

Yes, thanks for taking the questions. I'm going to be on the same couple of topics, but still it's on Agentic. If you kind of just take a look at some of the early more Agentic style use cases on Gemini AI mode and kind of observing some of the 3P integrations with Google Properties. Anything that surprised you or changed your view? I know we've talked a lot about commerce, but perhaps on like consumer productivity as well.

是的，謝謝讓我提問。我會圍繞相同的幾個主題，但仍然是關於 Agentic。如果你看看 Gemini AI 模式上一些早期、更偏 Agentic 風格的使用案例，並觀察 Google 產品（Properties）與一些第三方（3P）整合的情況。有沒有什麼讓你感到意外或改變你看法的地方？我知道我們談了很多商務，但也許也談談像消費者生產力（consumer productivity）之類的面向。

And then either Thomas or not, just on kind of cloud growth, is there anything you could share just around the rate of growth of Google Cloud, excluding the TPU system sales? And just how much of that's been kind of constrained by capacity? And as some of that repeat capacity comes in Q4, is it fair to kind of assume an acceleration of that Google Cloud business? Thanks.

另外，Thomas 或 Anat，關於雲端成長，你們是否能分享一下：若排除 TPU 系統銷售，Google Cloud 的成長率大概是什麼樣子？其中有多少是受到產能限制？而當部分新增產能在第四季到位時，是否可以合理假設 Google Cloud 業務會加速？謝謝。

Maybe I'll take the first part on. What we are building on these early agent style use cases, we are definitely seeing it. Just recently in the Gemini app, we have added a tab and where people actually type in agent queries, and so we can observe that.

我先回答第一部分。我們在這些早期代理式（agent-style）使用案例上所打造的內容，我們確實看到了。就在最近，在 Gemini 應用程式中，我們新增了一個分頁，讓人們實際輸入代理查詢（agent queries），因此我們可以觀察到這些行為。

Look, in all of this, users really want, they want to be delighted, so they effectively are looking for a high bar, what encourages repeat use is your ability to more handle these longer tasks and actually deliver a satisfactory outcome at the end.

你看，在這一切之中，使用者真的想要——他們想被驚喜、被取悅，所以他們實際上期待的是很高的門檻；能促使重複使用的，是你是否有能力處理這些更長的任務，並在最後真正交付令人滿意的結果。

Just like having seen this in other areas, including coding, et cetera. When you're doing it at 70%, it looks like it doesn't work until you cross a certain threshold and it really takes off. And so we are closely learned from all of it, identifying gaps, laying all the foundations I talked about earlier. And I'm confident it's an area where you will see a lot of progress across our surfaces as we go through later this year in '27.

就像我們在其他領域（包括寫程式等）也看過的一樣。當你只做到 70% 時，看起來好像不會成功，直到跨過某個門檻後才會真正起飛。因此，我們從所有這些經驗中密切學習，找出缺口，並把我先前提到的各項基礎都打好。我有信心，隨著我們在今年稍晚、一路到 2027 年的推進，你會在我們各個產品介面上看到大量進展。

The one thing as we transition to TPUs, one of the comments I would make on the TPU systems is look at semiconductor industry economies of scale matter. So I would look at our ability to do TPUs both for deploying in our data centers and TPU systems and other data centers as an ability to scale up the product which directly translates into better economics.

在我們轉向 TPU 的過程中，關於 TPU 系統我想補充的一點是：半導體產業的規模經濟很重要。因此，我會把我們打造 TPU（用於部署在自家資料中心，以及提供 TPU 系統給其他資料中心）的能力，視為擴大產品規模的能力，而這會直接轉化為更好的經濟效益。

To address the question on the rate of growth of cloud, Google Cloud segment delivered record growth in the third quarter, Cloud revenues up behind 3% to $30.7 billion. The strongest growth contributor was Google Cloud Platform with strength in both AI and core GCP. So GCP accounted for the vast majority of revenue generated in Q3. And Cloud revenue growth excluding the impact of GPU systems sales was again very strong.

針對雲端成長率的問題，Google Cloud 部門在第三季繳出創紀錄的成長，雲端營收成長 3%，達到 307 億美元。成長貢獻最大的來自 Google Cloud Platform，在 AI 與核心 GCP 兩方面都表現強勁。因此，GCP 佔了第三季所產生營收的絕大部分。

Demand was also strong across our product lines. So we see three things going on. One, as people use Gemini, we see revenue growth from Gemini itself. We also see that as people use Gemini, it drives growth in other parts of our portfolio. For instance, as they build agents, they use compute, perform a task, and so they buy compute.

我們各產品線的需求也都很強勁。所以我們看到三件事同時在發生。第一，隨著人們使用 Gemini，我們看到 Gemini 本身帶來的營收成長。我們也看到，當人們使用 Gemini 時，會帶動我們產品組合中其他部分的成長。例如，當他們建置代理（agents）時，會使用運算資源來執行任務，因此就會購買運算資源。

They use storage to store memories of the agent, database to serve up what's called context for the agent to reason on. And obviously, with all the cybersecurity threats, we are seeing strong growth in our cybersecurity portfolio. So we are confident about our growth and the diversity in our growth. We're also seeing new customers buy into GDP and accelerating growth in new customer adoptions. We also see that as a growth vector for us. And we are always at any time, we could obviously have a lot more demand with capacity, but we are managing all of these parameters and growth carefully.

他們會使用儲存來保存代理的記憶，使用資料庫來提供所謂的「情境」（context），讓代理用來推理。而且很明顯地，面對各種資安威脅，我們看到資安產品組合也有強勁成長。因此，我們對成長以及成長的多元性都很有信心。我們也看到新客戶採用 GCP，且新客戶導入的成長正在加速。我們也把這視為我們的成長動能之一。同時，我們在任何時候都可能因為產能而承接更多需求，但我們正謹慎管理所有這些參數與成長。

Michael Nathanson, Moffettnathanson.

Michael Nathanson，Moffettnathanson。

Thanks. I have two person here. In an early answer to Eric's question about CPU Cloud, we heard a willingness to work with some of your competitors or companies that we may view as competitors in model cloud. I wonder how you think about the philosophy here about working with people that maybe are competing with you and why this strategy is an opportunity, not a risk longer-term offense?

謝謝。我有兩個問題。在先前回答 Eric 關於 CPU Cloud 的問題時，我們聽到你們願意與一些競爭對手，或我們可能視為在模型雲端領域的競爭者合作。我想請問你如何看待與可能在跟你競爭的人合作的這種理念？為什麼這個策略在長期來看是機會，而不是風險或防守？

And then secondly on agentic consumer behavior, one of the debates is around safety and trust. Can you share with us what you've learned so far about safety and trust? Are those variables that matter to consumers? And do you think if those variables matter, what do you do it to protect consumer trust and privacy as you build out agentic consumer tools?

第二個問題是關於代理式（agentic）的消費者行為，其中一個爭論點在於安全與信任。你能否分享目前為止你們在安全與信任方面學到了什麼？這些變數對消費者重要嗎？如果重要的話，當你們打造代理式消費者工具時，會如何保護消費者的信任與隱私？

Thanks, Michael. Great questions. On the first one, look, I've long viewed what we build as categories of things which we build as products, and then there are categories of things which we build as platforms, right? And you need to have a clear view on which is which. And building things as platforms, you really work hard to make the platform great to make sure it works for everyone.

謝謝你，Michael。很好的問題。第一個問題，我一直以來都把我們打造的東西分成兩類：一類是我們作為產品來打造的，另一類是我們作為平台來打造的，對吧？你需要清楚分辨哪些屬於哪一類。而把東西作為平台來打造時，你就必須非常努力把平台做得很棒，確保它對所有人都能運作。

We have a long history of operating many platforms, be it a browser, be it operating system, be it an extraordinarily successful cloud platform. And in all of that, we worked hard to provide it to all players and be a great customer. You have seen us work on our models at other companies, including Apple to whom we compete in many areas.

我們長期以來營運許多平台，無論是瀏覽器、作業系統，或是極為成功的雲端平台。在這些平台上，我們都努力把它提供給所有參與者，並成為很好的合作夥伴。你也看過我們把自家的模型提供給其他公司使用，包括蘋果——我們在許多領域都與其競爭。

Similarly, we view our TPU systems as an important platform we are building. We want to be world-class in how we build it, means putting it against market forces, testing it and competing for people who would have the most demanding workloads on top of it. And I think that's what makes sure those teams are continuing to be at the frontier. So that's how we think about it. Over time, I think that always ends up in expanding opportunities and we've been able to do well taking that approach over a very long period of time.

同樣地，我們把 TPU 系統視為我們正在打造的一個重要平台。我們希望在打造它的方式上達到世界級水準，這意味著要把它放到市場力量下接受考驗，進行測試，並爭取那些在其上運行最嚴苛工作負載的客戶。我認為這能確保相關團隊持續站在最前沿。這就是我們的思考方式。隨著時間推移，我認為這樣做最終總會擴大機會，而我們長期以來採取這種做法也一直表現良好。

Second, on agentic behavior, I mean, to the earlier question, and I think I should emphasize. People really, when people are approaching these agent experiences, we clearly see people are concerned about safety and trust. I mean, it is a very high factor in how they want to use these things.

第二個，關於代理式行為，我想呼應先前的問題，也覺得我應該強調一下。當人們接觸這些代理體驗時，我們清楚看到大家確實在意安全與信任。這在他們想要如何使用這些工具的考量中，是非常重要的因素。

And over decades, the trust we have built by delivering these experiences in a safe way across many consumer products with scale in billions of users, as well as this are putting in end-to-end as we are building our models and agents, deploying both with extensive testing, the kind of real-time monitoring we do as we deploy these at scale. I think all of that will end up being a huge differentiator and we already hear about it from our users and I think they would vote with their feet as this frontier goes through.

而在過去數十年裡，我們透過以安全的方式在多個消費性產品中提供這些體驗、並在數十億使用者規模下運作，所建立的信任；再加上我們在打造模型與代理時採取端到端的做法，部署前進行大量測試，以及在大規模部署時所做的即時監控。我認為這些最終都會成為巨大的差異化因素；我們也已經從使用者那裡聽到這一點。我想隨著這個前沿領域的演進，使用者會用實際行動做出選擇。

Thank you. And that concludes our question-and-answer session for today. I'd like to turn the conference back over to Jim Friedland for any further remarks.

謝謝。以上是我們今天的問答環節。接下來我想把會議交回給 Jim Friedland，請他做進一步的結語。

Thanks everyone for joining us today. We look forward to speaking with you again on our fourth quarter 2026 call. Thank you and have a good evening.

謝謝各位今天加入我們。我們期待在 2026 年第四季電話會議上再次與各位交流。謝謝，祝各位晚安。

Thank you, everyone. This concludes today's conference call. Thank you for participating. You may now disconnect.

謝謝各位。今天的電話會議到此結束。感謝各位參與。您現在可以掛線。