Alphabet Inc. Class C parent company Alphabet (GOOGL.US) Q3 earnings call: TPU moves from Alphabet Inc. Class C cloud to customer data centers; 2027 will be an AI-driven "golden year of productivity"

Title context: Alphabet Inc. Class C parent company Alphabet (GOOGL.US) Q3 earnings call: TPU moves from Alphabet Inc. Class C cloud to customer data centers; 2027 will be an AI-driven "golden year of productivity" Text: The core thread presented by Alphabet Inc. Class C parent company Alphabet (GOOGL.US) on its Q3 2026 earnings call is that AI is accelerating the expansion of search usage, cloud revenue, and the TPU business, but there remains a time lag in converting search traffic into revenue, and compute buildout and significant TPU inventory investment are occurring before AI-related cash flow recovery. The company's Q3 revenue grew 25% year over year to US$127.6 billion; search query volume growth hit its highest rate in five years, but search advertising revenue growth slowed to 12%. Management explicitly acknowledged that, aside from currency effects and changes in advertising spend in certain verticals, the rapid growth of AI Mode queries combined with their currently lower monetization level is also one of the reasons for the slowdown in revenue growth at Alphabet Inc. Class C's traditional flagship businessthe search engine business. Management said on the earnings call that the company is expanding ads into more AI Overviews and AI Mode queries, advancing international markets and multi-turn conversational ads in the U.S., and exploring agentic commerce through the Universal Commerce Protocol, the Agent Payments Protocol, and the universal cart. Meanwhile, the Gemini first-party model API processes nearly 32 billion tokens per minute, up 40% quarter over quarter and more than four times the level a year earlier; the Gemini app has reached 1 billion users. Continued expansion of usage provides the foundation for future monetization, while improvements in ad coverage and business models will determine how this traffic converts into revenue. Cloud computing and TPU represent the strongest end of growth, but they also bring a new cash flow structure. The strongest business in Alphabet Inc. Class C's Q3 earnings seasonAlphabet Inc. Class C's cloud business Google Cloudsaw revenue grow 103% year over year to US$30.7 billion, with operating income reaching US$11.1 billion, more than triple the year-earlier level, and operating margin rising from 23.7% to 36.2%; backlog reached US$561 billion, with slightly more than 55% of that expected to be recognized as revenue over the next 24 months. Management emphasized that Google Cloud Platform contributed the vast majority of cloud revenue, and that growth remains extremely strong even excluding TPU system sales; enterprises' use of Gemini is also further driving demand for compute, storage, databases, and cybersecurity. TPU is extending from its own cloud services to customer data centers and partner channels, with system sales revenue expected to scale significantly in 2027. However, this business requires purchasing components and building inventory first, then recovering cash through delivery and invoicing; therefore, rapid growth in cloud business profit can coincide with pressure on the group's free cash flow. Q3 operating cash flow was US$42.3 billion, capital expenditure was US$51.4 billion, and free cash flow was negative US$9.1 billion; the increase in TPU inventory and the payment of a US$5.2 billion European Commission fine also further weighed on cash performance. Kurian specifically noted that all TPU-related inventory corresponds to signed contracts, and its financial impact is reflected as working capital tied up before contract delivery, rather than entirely being capital expenditure on the company's own infrastructure. 2027 is expected by management to be an AI-driven "golden year of productivity," and it will also be a year in which capital expenditure continues to grow significantly. The "golden year" emphasized by Alphabet Inc. Class C parent company CEO Pichai mainly refers to models such as Argon and Carbon unlocking productivity in coding, long-horizon tasks, and product innovation: internal teams have already achieved a 35% increase in code delivery and a 30% reduction in errors, and the company is also developing distilled model versions, secure runtime environments, and a tool-connection ecosystem to extend enterprise agent capabilities to a broader range of consumer products. Therefore, "harvest" is more accurately reflected as productivity gains, product iteration, and commercialization expansion, rather than management promising that free cash flow will fully turn positive or improve substantially next year. Alphabet maintained its 2026 capital expenditure guidance of US$195 billion to US$205 billion and expects a significant increase in 2027; infrastructure depreciation, energy, and operating costs will continue to affect the income statement, and expanded use of third-party compute in Q4 will also pressure near-term cloud margins. Management still expects that return on invested capital across the entire investment cycle will be significantly higher than the cost of capital, but the key transaction thread revealed by the call is already very clearAI demand and cloud revenue are materializing, and what the market needs to assess next is whether search monetization efficiency, TPU delivery collections, and continuously expanding investment can together form stronger cash generation capability. Below is the full transcript of the earnings call of Alphabet Inc. Class C parent company Alphabet (translated with the assistance of AI tools): Operator: Welcome, everyone. Thank you for standing by for Alphabet's Q3 2026 earnings call. At this time, all participants are in a listen-only mode. After management's remarks, we will conduct a question-and-answer session. (Operator instructions) Now, I will hand the call over to Jim Friedland, Head of Investor Relations, your host for today. Please go ahead. Jim Friedland: Thanks. Good afternoon, everyone, and welcome to Alphabet's Q3 2026 earnings call. Joining us today are Sundar Pichai, Philipp Schindler, Thomas Kurian, and Anat Ashkenazi. Now, I'll quickly cover the safe harbor for forward-looking statements. Some of the statements we make today about the company's business, operations, and financial performance may be forward-looking. Such statements are based on current expectations and assumptions and are subject to a variety of risks and uncertainties that could cause actual results to differ materially. Please refer to our 10-K and 10-Q reports, including the risk factors therein. We undertake no obligation to update any forward-looking statements. During this call, we will present financial measures on both a U.S. GAAP and non-U.S. GAAP basis. Reconciliations of non-U.S. GAAP measures to U.S. GAAP measures are included in the earnings press release issued today. The press release has been made available to the public through our Investor Relations website at abc.xyz/investor. Unless otherwise noted, our comments will be on a year-over-year basis. Now, I'll turn the call over to Sundar. Sundar Pichai: Thanks, Jim, and good afternoon, everyone. Q3 was a strong quarter for Alphabet. Our deep full-stack approach is driving growth across our businesses, and there is a lot more opportunity ahead. Let me give a quick overview of the quarter. Alphabet revenue grew 25% year over year, and we are now at a US$500 billion annualized revenue run rate. That is tremendous growth. Search momentum remains strong. Our AI investments are driving growth in usage and overall query volume, including commercial queries. In fact, query volume growth was the highest we have seen in the past five years. Google Cloud revenue grew 103%. AI infrastructure and solutions are the primary DRIVE of cloud growth, and customers see us as the leading provider in this area, so demand is very strong. We are working hard to bring more capacity online to meet this long-term demand. At the same time, demand for TPU systems from a broad customer base is very strong, and we expect that demand to continue as we scale. I am pleased to welcome Thomas to the call, and he will discuss cloud momentum further. On AI models, customer usage of our models is at unprecedented levels. Our first-party model APIs now process nearly 32 billion tokens per minute, up 40% from last quarter and more than four times the volume processed a year ago. With the launch of Gemini 4 Argon and Carbon announced today, we are excited to be at the forefront of technology. The Gemini app now has 1 billion users and is the fastest-growing product in Google's history. Finally, our global subscriptions across YouTube Music and Premium, including trialers, have surpassed 150 million. Today, I will discuss further our AI progress, our global consumer platforms led by Search and YouTube, and our long-term bets. First on AI, the most notable highlight is the Gemini 4 family of frontier models. Argon, which we launched two weeks ago, is seeing strong demand and usage. In fact, even with compute constraints, Argon ramped usage 3.5 times faster in its first week on our developer platform than Gemini 3. Argon performs exceptionally well in areas such as finance, law, cybersecurity, and science. Developers and cloud customers are already using it to untangle legacy codebases, accelerate scientific research, Siasun Robot&Automation technology, identify product improvement opportunities, and handle professional tasks requiring deep reasoning. Inside Google, in just the past few weeks, teams across Chrome, Ads, and Search have used Argon to significantly accelerate code writing and optimization, delivering 35% more code with 30% fewer errors. Argon is just the beginning of the Gemini 4 family. Today, we announced Carbon. This is our strongest frontier model, with state-of-the-art cybersecurity and long-horizon task capabilities. On the recently updated Artificial Analysis Intelligence Index, it ranks first with a score of 59. Carbon surpasses Argon in nearly every domain, including full-stack web development, multimodal understanding, complex knowledge work, and scientific reasoning. We are rolling out Carbon first to select trusted cyber defenders before expanding access; the U.S. government is also testing it. Beyond the Gemini 4 family, yesterday we also announced Genie 4, our most advanced world model yet, capable of simulating safe, realistic, interactive worlds with native audio support in real time. This is an important foundational building block for training Siasun Robot&Automation and agents. This rapid progress highlights the strong momentum in our model development. We have reached a stage where AI can help us improve the speed of training and release. I expect this process to continue accelerating over the next 12 months, driven by recursive self-improvement. We will also soon launch more models, including more efficient versions to help developers optimize speed and cost and deploy to billions of users. Next, Search. Search is bringing the benefits of AI to more people than any other product in the world. Just as with the shift to mobile more than a decade ago, we are focused first on building the best search user experience before advancing monetization. We are building a seamless, unified experience across AI Overviews, AI Mode, and traditional Search. People are increasingly using our AI features and searching in entirely new ways. Their queries are longer, more personalized, and more multimodal. Queries in AI Mode are on average three times longer than traditional search. Globally, one in every five AI Mode searches uses voice, images, or documents. We now have more than 50 billion visual searches per month. One area I am excited about is helping people actually get things done, whether booking, filling out forms, scheduling appointments, or shopping. As AI becomes more personalized and begins to take action on behalf of individuals and businesses, trust and safety matter more than ever. People already trust Google to rely on us for tasks in their daily lives, and our leadership in this area remains an important differentiator. We are focused first on getting these agent experiences right through Gemini agents for enterprises, because enterprises have extremely high requirements for security, protection, and governance. Next, we will bring these capabilities to Search and, of course, to the 1 billion users of the Gemini app. People are using the Gemini app not just to ask questions but to get things done, including complex multi-step agentic queries. The Gemini 4 family has made major progress here and already handles these complex tasks very well. Agents will become a unifying layer across our products, so you will see us extend these agent experiences from Search and the Gemini app further into Chrome, Android, and Workspace products. Next, YouTube. YouTube remains the No. 1 streaming platform in the U.S. by watch time and continues to lead in the living room. As mentioned, its subscriptions are also growing well. At last month's Made on YouTube event, we announced more than 30 products and features for creators and viewers. One good example is custom feeds: you simply choose your favorite topics and get a video homepage tailored around that topic. Philipp will discuss YouTube further. Next, our long-term bets. Waymo's trajectory is very strong. We have cumulatively completed more than 300 million miles of fully autonomous driving. Waymo currently operates in 15 major U.S. cities, has just opened to the public in San Antonio, San Diego, and Tampa, and plans to expand to Canada, Germany, Singapore, and Spain. Our other long-term projects are also making strong progress. Isomorphic is using Argon to support drug design research. We demonstrated magic state cultivation in quantum computing for the first time and sent TPUs into space through Project Suncatcher. So it was a very busy quarter, with much more to come. Thank you to our employees for their efforts this quarter. Philipp, over to you. Philipp Schindler: Thanks, Sundar, and good afternoon, everyone. I will focus on momentum in Search, YouTube, and partnerships. We continue to accelerate the deployment of Gemini models across our entire ads infrastructure to improve performance in the three areas I have highlighted before: ad quality, advertiser tools, and AI user experience. First, ad quality. Integrating Gemini models into our ads infrastructure is improving our ability to understand user intent and improve ad relevance. For example, in Q3 we made the biggest update to product listing ads in a decade, adding new AI features to help users make purchase decisions with greater confidence. We have already seen a 19% increase in satisfaction and more than a halving of dissatisfaction. We are also seeing that AI models can better predict clicks and conversions, driving stronger overall performance and deeper ad personalization. For example, through several innovations in personalized modeling, we delivered 18% incremental advertiser value for Discover ads in 2026 and increased user favorability by 8%. Overall, AI is accelerating our ad development process and driving launches across most of our products. Second, advertiser tools. We are ensuring that every advertiser, from global brands to small businesses, can seamlessly participate in our AI product interfaces. AI Max and Performance Max continue to see broad adoption. To give marketers more control, we are now introducing a new way for them to test different budgets and return-on-investment targets across multiple search campaigns in a single AI Max A/B test. To further improve performance, we are integrating AI creative tools directly into AI Max. In Q3 alone, Gemini models helped advertisers generate more than US$1.5 billion in campaign assets, up more than sevenfold year over year. For example, Telstra used AI Max to capture long and multilingual queries, increasing conversions by 52% and reducing cost per order by 36%, enabling it to reach customers at key moments no matter how they search. Third, AI user experience. Sundar mentioned the clear behavioral changes we are observing: users are asking entirely new categories of questions. We view Search holistically as a unified product spanning traditional results and AI search. In AI Mode, many queries are in the upper funnel, enabling us to capture new, higher-value commercial intent. In addition, most queries in AI Mode are incremental and will create more monetization opportunities over time. Also, our user research shows that people are receptive to ads in these interfaces and find them highly relevant. In AI Mode, click-through rates for paid promoted product listing items and organic product listing items are already at similar levels. Based on these combined results, we are now expanding high-quality ads in AI Overviews and AI Mode to more users and queries. This also includes extending our monetization coverage globally for the first time. Today, we announced the expansion of AI Mode ads to international markets, along with the launch of multi-turn conversational ads in AI Mode in the U.S. As always, we take a deliberate approach, prioritizing user experience and engagement first, and then scaling ad formats that have been proven to create returns for advertisers. That discipline provides a solid, durable foundation for our future. As Sundar said, agentic experiences across our services are still in the early stages, and we are excited about the commercial opportunity ahead. The Universal Commerce Protocol, or UCP, is being rapidly adopted by retailers. Today, UCP already supports hundreds of thousands of brands checking out directly across Google platforms, including Chewy, NIKE, Inc. Class B, Tapestry, and many others. We are combining it with the Agent Payments Protocol to build for the future of commerce. We are focused on two directions: first, unlocking agent capabilities that act on behalf of consumers, and second, designing practical tools that give shoppers confidence and control. To support the second direction, we launched the universal cart this quarter, aligned with how people actually shop across Google platforms today. This allows retailers to maintain their customer relationships while helping drive sales and reduce cart abandonment. Next, YouTube. We view YouTube as a flywheel of creators, viewers, and advertisers. First, creators. More than 20 million videos are uploaded every day. We are making content creation easier through new generative AI tools in YouTube Studio and a conversational editing assistant in Shorts and YouTube Create. Now, viewing. At Made on YouTube, we introduced a short-video series feature that lets creators organize their short videos into seasons and episodes. Continuous playback lets viewers seamlessly discover new short videos and keep watching. The third part is monetization, starting with ads. Our focus remains on creating value across the full marketing funnel. In Q3, living room viewing contributed more than one-third of total YouTube ad revenue. Small and medium advertisers continue to adopt Demand Gen to drive direct response performance. On YouTube, Wayfair's Demand Gen campaigns achieved a return on ad spend 2.3 times that of traditional video reach campaigns. Our YouTube creator partnership program connects brands with more than 3 million creators to help run effective campaigns. Brands such as Lexus continue to work with creators like Colin and Samir, increasing brand consideration by 11%. More than 1.3 million channels globally have joined YouTube Shopping. Building on this momentum, we recently announced that Amazon.com, Inc. is joining YouTube Shopping. U.S. creators can now tag Amazon.com, Inc. products directly in videos, Shorts, and live streams, driving direct conversions. Importantly, our subscription ecosystem continues to complement the ads business. Driven by YouTube Music Premium, subscriptions continue to grow faster than ads. YouTube Premium has more than 150 million subscribers. Through Premium and Premium Lite, viewers have different choices. As more people choose the Premium experience, viewers, creators, and our business all benefit. Subscription revenue helps support the creator ecosystem and moves the platform forward. Finally, strategic partnerships. Enterprises choose Google because we combine cloud, AI, ads, and consumer platforms to accelerate their transformation. We expanded our partnership with American Express Company to support the continued modernization of its contact centers and to collaborate on agentic commerce around Search and Wallet. Unity also signed an expanded multi-year cloud agreement to use our full-stack AI capabilities to power its next-generation engine. Our commercial priorities are very clear: improve advertiser ROI, systematically improve AI monetization efficiency, and expand ecosystem partnerships to support the next generation of discovery experiences. Finally, thank you to Google employees around the world for their outstanding execution, and thank you to our customers and partners for their continued trust. Thomas, over to you. Thomas Kurian: Our strong momentum comes from an integrated AI portfolio spanning chips, models, data, security, and agent platforms, all designed to work together. Our growth is highly diversified across product lines, industries, and customers. Google Cloud now has 18 product lines, each generating more than US$1 billion in annual revenue. Growth this quarter was driven by strong growth in Google Cloud Platform, with both AI and core GCP performing well. TPU system sales also made an important contribution. Over the past 12 months, more than 2,600 customers used at least 100 billion tokens on Gemini Enterprise, and nearly 75 customers used at least 10 trillion tokens. Nearly 90% of Fortune 100 companies use Gemini Enterprise. Customers that used Gemini Enterprise in Q3 2025 more than quadrupled their spend with us year over year. One important way customers use 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 Gemini agents, deeply integrated across our entire cloud portfolio. Through one agent, one API, and one prompt window, it can conduct research, answer questions, write and run code, and automate processes. It provides enterprises with strong security and cost controls and the broadest choice of frontier models. Customers include BNP Paribas, which is deploying Gemini Enterprise for more than 65,000 employees to simplify workflows while maintaining strict enterprise-grade security and data governance. Tata Steel deployed more than 300 specialized AI agents in just nine months to improve manufacturing safety, optimize supply chains, and cut customer service time by 50%. Ulta Beauty built the Ulta AI shopping assistant, which achieves a conversion rate three times that of the website's traditional browsing method. Gemini is also accelerating growth in our database, analytics, and cybersecurity products. Gemini can now help users find, analyze, and visualize information from structured and unstructured data. Gemini uses our knowledge catalog to find the right information sources and definitions, improving accuracy, reducing token costs, and making responses more reliable. Gemini uses our lakehouse platform to access information across multiple systems through zero-copy access. To support agentic analytics, BigQuery queries are more than 40% faster year over year, and query processing costs are more than 50% lower year over year. Customers include Deutsche Telekom, which used BigQuery, Cloud Storage, and the knowledge catalog to consolidate more than 40 legacy on-premises systems into a unified lakehouse platform. Etsy migrated 3 PB of data to a unified lakehouse platform. By using BigQuery and Iceberg, it now runs key data join operations up to 60% faster. Growing cybersecurity threats are also driving strong customer demand for our AI security platform. Our Gemini Flash cybersecurity model is a frontier model capable of automatically discovering vulnerabilities in 20 programming languages, with patching reliability 2.6 times that of other models. With the launch of Argon and Carbon, even better performance is ahead. Our security platform integrates threat intelligence, cyber response prioritization, automated scanning, code remediation, and monitoring. Gemini and Google threat intelligence autonomously analyze millions of external events every day with 98% accuracy. More than 95% of U.S. customers are adopting AI features, and growth continues. These security tools protect critical infrastructure, including financial institutions such as Banorte and Jack Henry, and telecom providers such as AT&T. All of this is supported by our leading AI infrastructure, spanning TPU systems, Axion CPUs, the latest NVIDIA Corporation GPUs, and Agent Compute Engine designed specifically for AI agents. TPU systems are purpose-built systems for AI, combining hardware, compute, storage, and networking with system components and software optimized for training and inference. Compared with the previous generation of TPU, TPU 8t delivers up to 2.7 times the training performance, and TPU 8i improves inference price-performance by 80%. To make it easy for developers to migrate workloads to TPU, we recently worked closely with Meta to launch TorchTPU, enabling users to train, serve, and deploy PyTorch models on TPU using tools they already know and rely on. Currently, nearly 100 customers use TPU, and we see strong and growing demand from many others. In addition to working directly and with Broadcom Inc. and Anthropic, our public customers also include Apple Inc., which is scaling pre-training of its foundation models on the latest generation of TPU. We are also working with new frontier model companies, including Mirendil and Physical Intelligence; with capital markets institutions, including Citadel Securities and Jump Trading; with universities such as Caltech; and with public sector service providers such as Cassava and Cerescale. Given continued strong and accelerating demand, we recently established TPU system delivery partnerships with Crux AI and Oracle Corporation, respectively, and more partnerships are in the works. Anat, over to you. Anat Ashkenazi: Thanks, Thomas. Unless otherwise noted, my comments will focus on year-over-year changes for Q3. I will start with Alphabet's overall results, then cover segment performance, and finally provide some commentary on the outlook for Q4 and full-year 2026. We had a strong Q3, delivering double-digit revenue growth for the 13th consecutive quarter. Consolidated revenue was US$127.6 billion, up 25% both as reported and on a constant-currency basis. Total cost of revenues was US$51.4 billion, up 24%. Traffic acquisition costs were US$16.4 billion, up 10%. Other cost of revenues was US$35.1 billion, up 32%, primarily driven by hardware costs related to TPU system sales, increased depreciation, content acquisition costs, and technical infrastructure operating costs. Total operating expenses grew 20% to US$35.6 billion. R&D expense grew 36%, primarily driven by compensation, mainly investment in AI talent, as well as increased depreciation. Sales and marketing expense grew 22% due to investment in Gemini and higher compensation. General and administrative expense declined 17%, primarily due to a comparison base effect from higher legal expenses in Q3 last year, partially offset by legal expenses this year. Operating income grew 30% to US$40.6 billion, with an operating margin of 31.8%. Other income and expense, net was negative US$8.2 billion, primarily due to an unrealized loss on our investment in SpaceX as its share price declined in Q3. Due to the decline in other income and expense, net, both net income and earnings per share declined. In Q3, we generated strong operating cash flow of US$42.3 billion, and operating cash flow over the past 12 months was US$179 billion. The increase in inventory needed for TPU system sales and the US$5.2 billion cash payment related to the 2018 European Commission fine after the appeal process adversely affected operating cash flow. Q3 capital expenditure was US$51.4 billion, the vast majority of which was for technical infrastructure to support our AI investments. Of the technical infrastructure investment in Q3, approximately 65% was related to servers and networking equipment, which typically have a useful life of six years. The remaining 35% was for data centers, with useful lives of seven to 40 years. In Q3, our free cash flow was negative US$9.1 billion, primarily driven by capital expenditure investment. The previously mentioned increase in inventory to support TPU system sales and the payment of the European Commission fine also affected Q3 free cash flow. Free cash flow over the past 12 months was US$19.7 billion. At quarter-end, we held US$265.1 billion in cash and marketable securities, including US$90.4 billion in marketable equity securities. Long-term debt was US$125.1 billion. We have also entered into multi-year agreements to secure technical infrastructure capacity, inventory, and lease and energy supply. At the end of Q3, our future purchase commitments were US$924 billion, and lease and other obligations not yet commenced were US$152 billion. These amounts reflect the maximum amount of contractual obligations, although some of these agreements have variable pricing and volume flexibility, so actual spending could be lower. As noted in the press release, the Board declared a quarterly cash dividend of US$0.22 per share on common stock, payable in December. Next, segment performance. Google Services revenue grew 11% to US$96.3 billion, reflecting strong growth in Search and subscriptions. Total advertising revenue grew 11%, with growth across all major verticals. As noted in our outlook last quarter, currency had a slight unfavorable effect on year-over-year growth, with a more significant effect on sequential performance. Google Search and other advertising revenue grew 12% to US$63.3 billion, with retail contributing the most, followed by finance. YouTube advertising revenue grew 9% to US$11.2 billion, with brand advertising contributing the most, particularly strong in the living room. Network advertising revenue was US$7.5 billion, up 1%. Subscriptions, platforms, and devices revenue grew 11% this quarter to US$14.3 billion, primarily driven by strong growth in YouTube subscriptions, especially YouTube Music and Premium, as well as strong demand for Google One's AI plans. Google Services operating income grew 17% to US$39.2 billion, with an operating margin of 40.7%. The Google Cloud segment delivered record growth in Q3. Cloud revenue grew 103% to US$30.7 billion, primarily driven by strong growth in GCP, with strength in both AI, including AI infrastructure and AI solutions, and core GCP. TPU system sales also made an important contribution. GCP contributed the vast majority of revenue in Q3. Excluding the impact of TPU systems, cloud revenue growth remained extremely strong. Cloud operating income was US$11.1 billion, more than tripling year over year. Operating margin rose to 36.2% from 23.7% in Q3 last year. The third-party transition compute discussed on our last call did not come online until the end of Q3, so its unfavorable impact on margins this quarter was limited. Google Cloud backlog reached US$561 billion in Q3, most of which relates to regular GCP contracts. We expect to recognize slightly more than 55% of total backlog as revenue over the next 24 months. Other Bets revenue was US$372 million, with an operating loss of US$2 billion, as we continue to scale Waymo and invest in important Other Bets projects. The operating loss for Alphabet-level activities was US$7.7 billion, primarily driven by shared AI R&D expenses. Now, I will provide the outlook for the remainder of 2026. First, revenue. At current spot rates, we expect the unfavorable impact of currency on consolidated revenue in Q4 to be slightly greater than last quarter. This impact will primarily be in Search and YouTube advertising. In Google Services, we will continue to face the high base created by the acceleration in Search performance that began in Q3 last year, which will pressure Q4 revenue growth. In Google Cloud, we expect TPU system sales to continue scaling in Q4 and to deliver significant revenue scaling in 2027. On cloud margins, given the supply-constrained environment, we began using third-party compute at the end of Q3 and will expand usage in Q4. As we use this compute, we expect near-term margin pressure. Next, investment. Our full-year 2026 capital expenditure guidance range of US$195 billion to US$205 billion remains unchanged. We continue to expect a significant increase in capital expenditure in 2027 and will provide more detail on the Q4 call. On expenses, the significant increase in technical infrastructure investment will continue to affect the income statement through higher depreciation expense and related data center operating costs, such as energy costs. Before we move to Q&A, I want to briefly discuss how we think about return on invested capital and free cash flow. We are uniquely positioned to monetize AI investment through multiple products for billions of users and enterprises. Our AI investment in Search is driving the highest query growth in five years. Year to date, Google Cloud has generated tens of billions of dollars in AI revenue; driven by demand for AI plans, Google One is our fastest-growing subscription product. Across the investment cycle, we expect to deliver attractive returns on invested capital that remain significantly above our cost of capital, and we expect to generate strong operating cash flow. As the TPU systems business scales, our significant investment in technical infrastructure and TPU inventory build will pressure free cash flow. Q3 was a strong quarter that highlighted our innovation and disciplined execution across our businesses, including the launch of Gemini 4. I want to take this opportunity to thank our employees for their contributions to the company's performance. Below is the Q&A session of the earnings call (AI-assisted summary and translation): Operator (Q&A session) Thank you. (Operator instructions) Our first question comes from Doug Anmuth of JPMorgan. Your line is open. QDoug Anmuth Thanks for taking the questions. I have one for Sundar and one for Philipp. Sundar, you said recently on a podcast that 2027 could be Google's golden age of productivity because of the new class of models and their integration with other products. I was hoping you could expand further and share how you see the opportunity ahead. And then, Philipp, query volume growth was the highest in five years, but search revenue growth has clearly slowed. Can you help us understand the difference? Is it simply because you are showing fewer ads in AI Overviews and AI Mode, and that portion of queries is now rising as a share of total queries? Thank you. ASundar Pichai Thanks, Doug. I would say that as models improve in long-horizon tasks, agentic coding, and so on, we were already beginning to see a step change internally when we used 3.8 Flash. But I do see a very significant inflection point now. Since we began deploying Argon internally, it has become one of the most important internal productivity levers we have seen in software engineering. We directly track high-quality code commits, commit velocity, and reductions in rollbacks. More importantly, people are taking on very large-scale problems; at the same time, we are beginning to see innovation at the product and consumer experience level after deploying these models. But all of this only started about four weeks ago. We are now beginning to deploy Carbon internally, and Carbon is another major step up from Argon. So as we roll out these models and provide more compute so that Google employees can use them broadly, I expectand this is what I was referring tothat after making those comments on the podcast, I am now even more excited. I think the coming year will be our golden year for improving productivity. So we are very excited about the innovation ahead. Philipp? APhilipp Schindler Okay, Doug, to answer the second part of your question. Q3 search revenue grew 12% year over year to US$63 billion. If you look at the change in growth from Q2 to Q3, there are several drivers. First, currency, which was the single largest factor in the slowdown in search growth compared with Q2. Second, changes in certain verticals. In retail, some of our Asia-Pacific advertisers were affected by the EU's removal of the de minimis import duty exemption. In media and entertainment, Q2 was exceptionally strong due to the FIFA World Cup, after which spending normalized. The World Cup had more matches in June, so more ad budget was actually allocated to June than to July. To your next point, it is worth noting that the strong growth in AI Mode queries also had an effect, because the monetization level we were seeing in AI Mode at the time was lower. To be clear, we are very confident in the trajectory here. Ad experiments are already showing encouraging results, and we are now expanding ads in AI Overviews and AI Mode to more users and more queries, including the international markets I mentioned and multi-turn conversational ads in the U.S. QDoug Anmuth Thank you. Operator Our next question comes from Brian Nowak of Morgan Stanley. Your line is open. QBrian Nowak Thanks for taking my questions. I have two. Starting with agents, you launched Spark to some users during the year. I know, Sundar, you have talked a lot about plans to bring more agent products to a broader user base. Can you help us understand the key technical hurdles you need to overcome to bring agent products to all users? Is it model efficiency? Is it compute? Or is the technology not yet ready? Technically, what exactly do we still need to do? And second, following up on Doug's question about Search, Anat, can you help us better understand the comparison factors you mentioned affecting Search? How should we think about those comparisons as we move into Q4 and Q1? Thank you. ASundar Pichai Thanks, thanks Brian. First, agents are an exciting and dynamic area. When I look at our new models and see the complex tasks they can autonomously take on and execute at very high quality, it is clear that the agent capabilities we have long envisioned will become reality. So I am very much looking forward to bringing the next generation of agent experiences to users. To your question, I would make a few points. The Gemini 4 model family is very important. We are developing distilled versions of the Carbon model so that we can deploy efficiently to a billion users in terms of compute, latency, and quality. So this is an important area of investment. To deploy at scale, you clearly need more than a capable model; you need a robust agent execution framework and a highly secure runtime environment, which is where GCP comes in. We need to provide a secure runtime environment available around the clock for every user in the cloud, as well as an appropriate ecosystem of skills and connectors. All of these are foundational. We have been working on them and decided to launch first for enterprises, which is the Gemini agents for enterprises we announced, because enterprises have the highest requirements for security governance and control. It has been well received, and we are expanding deployment. Later this year, we will bring the same experience to consumers, clearly starting with the 1 billion users in the Gemini app. We are already seeing that how users use the app is shifting from chat to tasks and agentic experiences. Users' trust in our products and all their context information are already there, which I think creates favorable conditions for us. But as a company, we are really focused on delivering this unified experience across products, and you should see a lot of progress later this year. AAnat Ashkenazi Thank you. On the second question, which relates to my comments, I think you are referring to the Search outlook. I mentioned several factors. One is the sequential currency headwind, which also occurred this quarter versus last quarter. So at current spot rates, we expect that to create some headwind for Q4. Then looking at year-over-year comparisons, this quarter we began to face the high base created by the acceleration in Search growth in 2025, that is, in the second half of 2025. The comparison base relevant to Q3 included about three points of growth acceleration; we expect this to continue in Q4, facing a comparison base impact of about a few points from the acceleration in late 2025. QBrian Nowak Great, thank you both. Operator Our next question comes from Eric Sheridan of Goldman Sachs Group, Inc. Your line is open. Q