Google conference call: Full-stack AI strategy drives comprehensive business growth, with monthly token processing volume doubling and visual search growing by 70%

Google's Q2 earnings report exceeded expectations, with AI business advancing comprehensively, processing volume doubling to 98 trillion per month. The search business showed impressive growth, with AI overview driving over 10% growth in queries globally. Capital expenditures significantly increased to $85 billion for AI infrastructure construction. On July 23, Google announced its Q2 earnings report, showing a 14% year-on-year revenue growth to $96.4 billion. Cloud computing became the biggest beneficiary of the AI boom, with revenue reaching $13.6 billion, a 32% year-on-year increase, accelerating from 28% in the first quarter. Additionally, despite facing competitive threats from AI products like OpenAI's ChatGPT and Perplexity, Google's core search business demonstrated resilience, with search-driven ad revenue reaching $54.2 billion, a 12% year-on-year increase. Google raised its annual capital expenditure forecast by 13% to approximately $85 billion, a significant increase from $52.5 billion in 2024. After the earnings report was released, Google CEO Sundar Pichai, CFO Anat Ashkenazi, CBO Philipp Schindler, and Senior Director of Investor Relations Jim Friedland attended a conference call to answer questions regarding AI infrastructure, increased capital expenditures, cloud business outlook, generative AI, the AI talent competition, and the impact of AI overview on ad displays on search results pages. CEO Sundar Pichai stated during the earnings call, "Q2 was an outstanding quarter for us, with strong growth across the company. As you saw at Google I/O, we are at the forefront of AI and advancing at an incredible pace. AI is positively impacting every aspect of our business, driving strong momentum." "From everything we see right now, people are excited about AI, and they are widely adopting it across our various products. For me, seeing multimodal applications and how people are changing their behavior to naturally incorporate images into their interactions with Google, whether through Google Lens or Circle to Search, shows that these features have been well accepted as part of their daily experience." The company launched the "AI Mode" search feature in May to answer user queries in a conversational interface. Pichai also emphasized that AI features have not weakened the search business. MoffettNathanson analyst Michael Nathanson stated that Google executives effectively alleviated shareholders' concerns about the future of the search business during the earnings call. He said: "What people are worried about is the mathematical logic of search; won't AI negatively impact search? They emphasized multiple times during the call that 'this is not what we are seeing.'" Summary of key points from the earnings call of Google's parent company Alphabet: 1. AI business fully advanced, processing volume doubled: "At the I/O conference in May, we announced that our services process 480 trillion tokens per month. Since then, we have doubled that number, currently processing over 980 trillion tokens per month, a remarkable growth. The current monthly active users of the Gemini application have exceeded 450 million, and we continue to see strong growth and user engagement, with daily request volume increasing by over 50% compared to the first quarter." 2. Search business shows impressive growth, AI overview drives query growth: "This quarter, Search achieved double-digit revenue growth. The AI overview appeared in the query types of this feature, driving over 10% query growth globally. Search and other revenues increased by 12%, reaching $54.2 billion, with strong performance across all verticals, and retail and financial services contributing the most." 3. Significant increase in capital expenditure, raised full-year expectations: "Given the strong demand for our cloud products and services, we now expect capital expenditures for 2025 to be around $85 billion, up from the previous estimate of $75 billion. Our updated expectations reflect additional investments in servers, changes in server delivery timing, and accelerated data center construction progress, primarily to meet the needs of cloud customers." 4. Supply and demand remain tight, investment will continue until 2026: "Although we have been working to increase capacity and have accelerated server deployment, we still expect that the demand-supply gap will remain tight until 2026. Looking ahead to 2026, due to sustained customer demand and business expansion opportunities across the company, we expect capital expenditures to further increase." 5. AI-driven advertising innovation: "Search advertisers using AI Max typically achieve a 14% increase in conversion. In media buying, Smart Bidding Exploration is the biggest update to bidding strategies in a decade... Campaigns using Smart Bidding Exploration average a 19% increase in conversions." 6. YouTube performs strongly, short video revenue on par with traditional videos: "In the U.S., the hourly viewing revenue of Shorts has now matched that of traditional YouTube video ads. In some countries, it has even surpassed the revenue rate of traditional videos. YouTube ad revenue increased by 13%, reaching $9.8 billion, primarily driven by direct response ads, followed by brand ads." 7. Outstanding performance in large client contracts: "First, the number of contracts exceeding $250 million has doubled year-on-year; second, in the first half of 2025, the number of contracts over $1 billion we signed is on par with the entire year of 2024." 8. Growth of Visual Search: "Google Lens search is one of the fastest-growing search types, with a 70% increase since the same period last year. Most Lens searches are new traffic, and we are also seeing good growth in shopping searches using Lens." 9. Waymo Continues to Expand Service Scale: Last month, Waymo launched services in Atlanta, more than doubled its service area in Austin, and expanded its service area in Los Angeles and the San Francisco Bay Area by about 50%. Waymo also introduced a teen account, initially serving passengers aged 14-17 in Phoenix. The Waymo Driver has now accumulated over 100 million miles of autonomous driving on public roads. The team will conduct tests in more than 10 cities this year, including New York and Philadelphia, with hopes to launch services in all 10 cities in the future. Strong Performance Growth Driven by AI In the second quarter, Alphabet achieved outstanding financial performance, with consolidated total revenue reaching $96.4 billion, a year-on-year increase of 14%. This growth is primarily attributed to the deep application of AI technology across various business segments. Search business revenue grew by 12% to $54.2 billion, YouTube ad revenue increased by 13% to $9.8 billion, and Google Cloud business achieved a strong growth of 32%, reaching $13.6 billion. AI is actively impacting every aspect of the company's business, from the increase in search query volume to the surge in demand for cloud services, reflecting the tremendous driving force of AI technology. The company's AI summary feature has over 20 million monthly active users, covering more than 200 countries and regions, supporting 40 languages, while the monthly active users of the Gemini application exceed 450 million. Technological Leadership of Full-Stack AI Strategy Google has demonstrated its comprehensive technological strength in the AI field, covering three core areas: AI infrastructure, world-class research, and models and tools. In terms of infrastructure, the company operates globally leading AI-optimized data centers and cloud regional networks, providing the industry's most extensive TPU and GPU combinations. In the research aspect, the Gemini 2.5 series hybrid reasoning models have shown industry-leading performance in almost all major benchmark tests, with over 100 million developers currently using Gemini to build products. In terms of product application, it processes over 980 trillion tokens monthly, doubling from 480 trillion in May. This full-stack layout from underlying infrastructure to upper-level applications ensures Google's differentiated advantage in the AI competition. AI Innovation in Search Experience and User Growth AI is fundamentally changing user search behavior and experience. The AI summary feature has driven over a 10% increase in queries globally for the types of queries where this feature appears, and this growth continues to rise The monthly active users of AI models in the United States and India have exceeded 100 million, providing an excellent experience for users to handle longer and more complex issues. The use of multimodal search, especially Google Lens, has increased by 70%, and Circle to Search is running on over 300 million Android devices. These AI-driven new search experiences not only enhance user satisfaction but also significantly increase search frequency, particularly among younger user groups. Users are beginning to realize that search can meet more needs, making search behavior more frequent and in-depth. Rapid Expansion of Cloud Business and Future Investment Plans Google Cloud has shown strong growth momentum, with an annual revenue run rate now exceeding $50 billion, and operating profit margin has significantly increased from 11.3% to 20.7%. The number of large orders exceeding $250 million has doubled year-on-year, and the number of large orders over $1 billion signed in the first half of 2025 is on par with the entire year of 2024. More than 85,000 companies are using Gemini to build services, driving a 35-fold year-on-year increase in Gemini usage. To meet the continuously growing customer demand, the company has raised its 2025 capital expenditure expectation from $75 billion to $85 billion, with about two-thirds allocated for server investments. In the face of a tight supply-demand market environment, Google is accelerating data center construction and server deployment, and this supply tightness is expected to continue until 2026, with capital expenditures in 2026 also expected to increase further. The full AI translation of Google's conference call is as follows: Alphabet Q2 2025 Earnings Call Date: July 23, 2025 Company Name: Alphabet Event Description: Q2 2025 Earnings Call Source: Alphabet Moderator: Welcome, everyone. Thank you for your patience in attending the Alphabet Q2 2025 earnings call. Currently, all participants are in listen-only mode. After the presentation, there will be a Q&A session. (Operational Prompt) Now, I will hand the call over to today's speaker, Jim Friedland, Head of Investor Relations. Please go ahead. Jim Friedland, Senior Director, Investor Relations: Thank you, everyone. Good afternoon, and welcome to the Alphabet Q2 2025 earnings call. Joining us today are Sundar Pichai, Philip Schindler, and Anat Ashkenazi. I will quickly introduce the safe harbor statement. Some statements we make today regarding the company's business, operations, and financial performance may be considered forward-looking statements. Such statements are based on current expectations and assumptions and are subject to various risks and uncertainties. Actual results may differ materially from these statements Please refer to our 10-K and 10-Q forms, particularly the "Risk Factors" section. We are not obligated to update any forward-looking statements. In this conference call, we will provide U.S. Generally Accepted Accounting Principles (GAAP) and non-GAAP financial metrics. Reconciliation information between non-GAAP and GAAP is included in today's earnings press release, which has been published and made public through our investor relations website (abc.xyz/investor). Unless otherwise noted, our comments are based on year-over-year comparisons. Now, I will turn the meeting over to Sundar. Sundar Pichai, CEO: Thank you, Jim. Good afternoon, everyone. The second quarter has been an outstanding quarter for us, with strong growth across the company. As you saw at the Google I/O conference, we are at the forefront of artificial intelligence and advancing at an incredible pace. AI is actively impacting every aspect of our business, driving strong momentum. This quarter, Search achieved double-digit revenue growth. Our new search features are performing well. The AI model has been launched in the U.S. and India and is progressing smoothly, with monthly active users of the AI summary feature exceeding 20 million, covering over 200 countries and regions and supporting 40 languages. I will elaborate on Search later. We continue to perform strongly in YouTube and subscription businesses, both of which are high-growth areas. In the U.S., the revenue per hour of Shorts has already matched that of traditional YouTube video ads. In some countries, it has even surpassed the revenue rates of traditional videos. The Cloud business is also performing exceptionally well, with strong growth in revenue, order backlog, and profitability. Its annual revenue run rate has now exceeded $50 billion. We are seeing significant growth in customer demand for our comprehensive AI product portfolio. Of course, all of this is thanks to our long-term investments in AI, including a differentiated full-stack strategy. This encompasses AI infrastructure, world-class research, models and tools, as well as our products and platforms dedicated to bringing AI to users worldwide. Before moving on to quarterly highlights, I will briefly discuss our AI stack. First is AI infrastructure. We operate the world's leading AI-optimized data centers and cloud regional networks. We also offer the industry's broadest combination of TPUs and GPUs, along with supporting storage and software. This is why nearly all generative AI unicorns use Google Cloud. Consequently, more and more companies, including leading AI research labs like Safe Superintelligence and Physical Intelligence, specifically choose to use TPUs. Our investments in AI infrastructure are crucial to meeting the growing demands of cloud customers Next is world-class AI research, including models and tools. We continue to expand our Gemini 2.5 series hybrid reasoning models, which have demonstrated industry-leading performance in almost all major benchmarks. In addition to optimizing our popular flagship model Flash, we have also launched the extremely fast Flashlight version. We achieved gold medal-level performance in the International Mathematical Olympiad by combining the advanced version of Gemini with Deep Think. We can't wait to bring Deep Think to a wider audience. We currently have one of the best models on the market across various price ranges. Our 2.5 model has become a catalyst for growth, with over "0 million" (the original text may have meant "100 million") developers using Gemini to build products. I also want to mention Veo 3, our advanced video generation model. It has gone viral online, with users sharing their generated clips in the Gemini app, paired with our new AI filmmaking tool Flow. Since May, over 70 million videos have been generated. We recently introduced a feature in the Gemini app that converts photos into videos, which has received excellent feedback from users. This feature is rolling out to Google Photos users today. Third, our products and platforms. We bring AI to all users and partners through multiple interfaces such as Workspace and Chrome. The usage and growth are remarkable. At the I/O conference in May, we announced that our services process 480 trillion tokens per month. Since then, we have doubled that number, currently processing over 980 trillion tokens per month, with impressive growth. The current monthly active users of the Gemini app have exceeded 450 million, and we continue to see strong growth and user engagement, with daily request volume increasing by over 50% compared to the first quarter. In June alone, over 50 million people used the AI meeting minutes feature in Google Meet. With the support of Veo 3, our new short video product Google Vids launched in Workspace has nearly 1 million monthly active users. This month, we announced new Android and AI features at the Samsung Galaxy Unpacked event, which will be applied to Samsung's latest devices. We are also very pleased with the growth of our subscription business, partly due to the launch of our Google AI Pro and Ultra subscription plans. Now, let's introduce some key highlights from this quarter's Search, Cloud, YouTube, and Waymo First of all, this is an incredibly exciting moment for Search. We see that AI is driving the expansion of search methods, making the ways people search for and obtain information more diverse, allowing Google to respond to entirely new types of questions. The overall query volume and commercial query volume for Search continue to achieve year-on-year growth, with our new AI experiences making a significant contribution to this growth. We have also observed that as users realize that Search can meet more of their needs, their search behavior becomes more frequent. This is particularly evident among younger user groups. Let me elaborate on our new search experience. We know that AI Overviews are very popular, as they are now driving over 10% query growth globally among the types of queries where this feature appears. Moreover, this growth continues to rise. AI Overviews are now powered by Gemini 2.5, providing the industry's fastest AI response speed. We have also seen strong growth in multimodal search, especially when Lens or Circle to Search is used in conjunction with AI Overviews. This growth is most pronounced among younger users. Our brand new end-to-end AI search experience, AI Mode, continues to receive positive feedback, particularly excelling in handling longer and more complex questions. It is still being gradually rolled out, but the monthly active users in the United States and India have exceeded 100 million. We plan to continuously optimize the AI Mode experience, quickly launching excellent new features, including our advanced research tool Deep Search and more personalized response capabilities. Next is Google Cloud. We see strong customer demand, primarily due to our product differentiation advantages and comprehensive AI product portfolio. Four data points illustrate this: The number of large orders exceeding $250 million has doubled year-on-year; In the first half of 2025, the number of large orders over $1 billion we signed is on par with the entire year of 2024; The number of new GCP (Google Cloud Platform) customers has increased by nearly 28% quarter-on-quarter; Over 85,000 enterprises—including LVMH, Salesforce, and Singapore's DBS Bank—are now using Gemini to build services, driving a 35-fold year-on-year increase in Gemini usage. Our models rely on AI infrastructure to provide services, which has demonstrated industry-leading performance and cost efficiency in both training and inference. In addition to AI accelerators, we have also launched innovative features in storage, including Anywhere Cache, which can reduce inference latency by up to 70%, and Rapid Storage, which improves latency by five times compared to leading hyperscale service providers In addition, we have optimized AI software packages, including PyTorch and JAX, and provided comprehensive open-source support for various AI training and service needs. We have also deeply integrated AI Agent into our Cloud products. Wayfair (U.S. home e-commerce) is leveraging our integrated AI database to streamline data pipelines and provide customers with a more personalized experience. Vantel is utilizing our Gemini Power data agent and BigQuery to review and process product feedback more quickly. Target is using our Gemini Power threat intelligence and security operations agent to enhance cybersecurity levels. Capgemini is using our AI software engineering agent to deliver higher quality software faster through task automation from code generation to testing. BBVA (Banco Bilbao Vizcaya Argentaria) stated that Gemini and Google Workspace are helping employees save nearly 3 hours a week on repetitive tasks. BBVA is currently rolling out this technology to its 100,000 employees worldwide. We are also focused on building a thriving AI agent ecosystem. We launched an open-source agent development kit, which has been downloaded over 1 million times in less than four months. We also launched Agentspace, an open and interoperable enterprise chat search and agent platform. Gordon Food Service is introducing Agentspace for its U.S. employees, making the decision-making process more efficient and high-quality. Before its official launch, Agentspace has already booked over 1 million subscription seats. Now we turn to YouTube. Nielsen data shows that YouTube has led in streaming watch time in the U.S. for two consecutive years. The generation of users who grew up with YouTube on mobile devices is now increasingly watching their favorite creators and content on television. This includes millions of sports fans. Globally, they watch over 40 million hours of sports content on YouTube each year. In September, we will livestream the first NFL Friday game of the season, taking place in Brazil. From sports to Shorts, we currently have an average daily view count of over 200 million on YouTube Shorts AI is helping us improve recommendation systems and auto-dubbing, significantly expanding the potential audience for creators and brands, leading to better returns. Today, we are launching a whole new batch of AI tools on YouTube Shorts to support creators. Finally, YouTube continues to diversify its subscription options, recently expanding its Premium Light service to 15 new countries, with more countries to come in the future. One last point, Waymo continues to scale its services to safely provide rides to more passengers in more areas. Last month, Waymo launched services in Atlanta, more than doubled its service area in Austin, and expanded its service area in Los Angeles and the San Francisco Bay Area by about 50%. Waymo also introduced a teen account, initially serving passengers aged 14-17 in Phoenix. Overall, this is a very strong momentum. The Waymo Driver has now accumulated over 100 million miles of autonomous driving on public roads. The team will be testing in more than 10 cities this year, including New York and Philadelphia, and we hope to launch services in all 10 cities in the future. As I said, this is a standout quarter. As always, a heartfelt thank you to our employees and partners for all their efforts in the second quarter. Philipp, I’ll hand it over to you. Philipp Schindler, Senior Vice President and Chief Business Officer of Google: Thank you, Sundar, hello everyone. I will quickly introduce the performance of Google Services this quarter, and then I will elaborate on our significant progress in search, advertising, YouTube, and partnerships. This quarter, Google Services revenue was $83 billion, a year-over-year increase of 12%, primarily driven by strong growth in search and YouTube, partially offset by a year-over-year decline in network revenue. To better explain our performance, the 12% growth in search and other revenues was driven by growth across all verticals, with retail and financial services contributing the most. YouTube also performed similarly across various verticals. Its advertising revenue grew 13% year-over-year, primarily driven by direct response ads, followed by brand ads. Starting with search and other revenues, this segment brought in over $54 billion this quarter Transformations like artificial intelligence are driving our industry forward. Gemini's native multimodal capabilities are helping to bring the offline audio and visual world back online, creating numerous opportunities for search. Let me share a few examples: For instance, visual search. Google Lens search is one of the fastest-growing search types, with a 70% increase since the same time last year. Most Lens searches are new traffic, and we are also seeing good growth in shopping searches conducted using Lens. You can obviously elevate this feature to a higher level, such as evolving from image search to video-based functionalities like Search Live. Then there's Circle to Search, which is currently running on over 300 million Android devices. We are continuously adding new features to help users explore complex topics and ask follow-up questions without switching apps. For example, gamers can now use Circle to Search while playing mobile games to see AI summaries or answers. Just last week, we introduced an AI-driven proactive agent feature directly in search for all users in the U.S., which can call local businesses. Finally, there's the shopping experience. In the second quarter, we launched a virtual try-on experience for Search Labs users in the U.S. Now users can virtually try on billions of clothing items. Initial feedback and engagement have been very positive, especially among Gen Z users. We are about to roll out this feature to all users in the U.S. All these innovations are opening up new ways for people to use technology, bringing the offline world online in unprecedented ways. Coupled with our outstanding AI translation capabilities, imagine the potential. People can access more content in their own language, reaching more customers, whether they are large or small, international or local businesses. I am very excited about the fusion of these elements and the future opportunities in search. Next, let's talk about advertising. Our strategy is to reshape the entire marketing process with AI, bringing value to our clients and our business. Last quarter, we launched AI Max for Search, a new suite of AI features integrated into existing search ads. Advertisers using AI Max typically see a 14% increase in conversion. In media buying, Smart Bidding Exploration is the biggest update to bidding strategies in a decade, helping advertisers bid more frequently on less obvious but potentially higher-value search terms, thereby enhancing ad effectiveness Campaigns using Smart Bidding Exploration can achieve an average conversion growth of 19%. Demand Gen ads continue to drive revenue growth and deliver measurable results for clients. For example, Depop (Etsy's clothing resale platform) used a Demand Gen campaign limited to Shorts to attract new customers to its website. The brand lift effect from Shorts ads exceeded the benchmark by 80%, and the click-through rate was twice the benchmark. In terms of creativity, we launched Asset Studio, leveraging our latest models to help businesses of all sizes generate creative assets. Small businesses can access high-quality ad materials and achieve rapid deployment, while large enterprises can move faster from proof of concept to launch and scale at a lower cost. Currently, over 2 million advertisers are using Google's AI-driven creative generation tools for advertising, a 50% increase compared to the same period last year. Now, talking about YouTube, we see strong and continuous growth in ad revenue, primarily driven by direct response ads, followed by brand ads. YouTube creators are closely connected to global cultural trends and enjoy high trust from audiences. In the BrandConnect program, we launched the Creator Partnership Hub to help brands more easily find the right creators and seize cultural opportunities. We also introduced Veo 3, photo-to-video features, and generative effects on Shorts, making content creation simpler while providing more untapped creative directions. We see an increase in the number and price of Shorts ads, especially in developed markets. Due to the feed nature of Shorts, there are more ad opportunities on average per piece of content. This growth is also benefiting from ad formats tailored specifically for Shorts, AI-driven creative adjustment tools, more precise ad targeting, and increased audience engagement rates. McDonald's USA leveraged the influence of YouTube creators to promote the Minecraft co-branded meal, increasing ad reach through YouTube Shorts collaborative ads, with a view-through rate (VTR) 3.3 times the industry benchmark. Finally, looking at CTV (Connected TV), its growth momentum remains strong. According to Nielsen's Gauge report, YouTube has been the number one platform for streaming watch time in the U.S. for two consecutive years, reaching a record 12.8% of total TV watch time by June 2025 In the past 12 months, YouTube ads viewed through CTV screens have generated over 1 billion conversions. CTV shopping ads have driven retail growth, allowing viewers to shop directly through QR codes, helping us better capture direct marketing opportunities. As usual, I will conclude by summarizing our momentum in partnerships, as clients increasingly recognize Google's strength and breadth in leveraging AI to help businesses transform. For example, our new partnership with PayPal will enhance the digital shopping experience for its merchants and customers. PayPal will expand its applications of AI-driven recommendations, transaction processing, and enhanced security on Google Cloud. This partnership also expands the availability and coverage of PayPal's payment services and features across multiple Google products. Finally, I want to thank all Google employees for their contributions and commitment to our success, and also thank our customers and partners for their continued trust and support. Anat, it's over to you. Anat Ashkenazi, Senior Vice President and Chief Financial Officer: Thank you, Philipp. My remarks will focus on year-over-year comparisons for the second quarter, unless otherwise noted. I will start with the overall performance at the Alphabet level, then discuss the performance of each business segment, and finally make some comments on the outlook for the second half of 2025. We performed steadily in the second quarter. Consolidated total revenue was $96.4 billion, a year-over-year increase of 14%, and a 13% increase at constant currency. Among these, search and YouTube ads, subscription platforms and hardware devices, as well as Google Cloud all achieved double-digit revenue growth, reflecting the strong momentum of the company's overall business. Total revenue costs were $39 billion, a year-over-year increase of 10%. Among these, technology costs were $14.7 billion, a year-over-year increase of 10%; other revenue costs were $24.3 billion, a year-over-year increase of 10%, with the main growth source being YouTube's content procurement costs, followed by depreciation expenses. Total operating expenses were $26.1 billion, a year-over-year increase of 20%. The growth was mainly due to legal and other expenses, with the increase reflecting a $1.4 billion provision related to preliminary settlements of certain legal matters. R&D spending grew by 16%, mainly due to increases in compensation and depreciation expenses. Sales and marketing expenses grew by 5%, primarily reflected in increased advertising and promotional spending. This quarter, operating income reached $31.3 billion, a year-over-year increase of 14%, with an operating profit margin of 32.4%. The profit margin benefited from strong revenue growth and continued cost control, but was partially offset by the aforementioned legal expenses and increased depreciation Net profit reached USD 28.2 billion, a year-on-year increase of 19%, with earnings per share of USD 2.31, up 22% year-on-year. Free cash flow for the quarter was USD 5.3 billion, with total free cash flow over the past 12 months reaching USD 66.7 billion. The free cash flow for this quarter was impacted by a significant increase in capital expenditures in the second quarter and the payment of federal tax cash for the first and second quarters. As of the end of this quarter, the company held cash and cash equivalents of USD 95 billion. Next is the performance of each segment: Google Services revenue grew by 12% to USD 82.5 billion, reflecting strong performance in search, YouTube advertising, and subscription services. Among them, search and other revenues increased by 12% to USD 54.2 billion, with strong performance across all verticals, particularly in retail and financial services. YouTube advertising revenue increased by 13% to USD 9.8 billion, primarily driven by direct response ads, followed by brand ads. Network advertising revenue was USD 7.4 billion, a year-on-year decrease of 1%. Subscription platforms and device business revenue grew by 20% to USD 11.2 billion, mainly due to growth in YouTube subscription services, followed by growth in Google One. Google Services operating income grew by 11% to USD 33.1 billion, with an operating margin year-on-year remaining flat at approximately 40.1%. This level of profitability benefited from revenue growth and ongoing cost optimization, but was also impacted by legal expenses. Turning to Google Cloud: This segment performed very strongly this quarter. Revenue grew by 32% year-on-year to USD 13.6 billion, significantly outpacing overall cloud business growth, while Google Workspace revenue also saw strong growth due to an increase in seat count and unit price. Google Cloud operating income reached USD 2.8 billion, with the operating margin increasing from 11.3% to 20.7%. The rise in profit margin came from revenue growth and ongoing cost efficiency improvements, although partially offset by higher costs of technology infrastructure usage (including depreciation). As we increase our investment in AI, we are also working to enhance productivity and cost-effectiveness to offset the cost pressures from depreciation. The backlog for Cloud business grew by 18% quarter-on-quarter and 38% year-on-year, reaching USD 106 billion by the end of this quarter. This growth reflects strong demand for products and services from both new and existing customers. As Sundar mentioned, we signed multiple contracts exceeding USD 1 billion in the first half of the year. Other Bets: Revenue for the second quarter was USD 373 million, with an operating loss of USD 1.2 billion In Other Bets, we are investing more resources into businesses with the potential to create additional value, such as Waymo. Next, let's talk about capital expenditures (CapEx): This quarter's CapEx was $22.4 billion, with approximately two-thirds allocated to servers and one-third to data centers and networking equipment, with nearly all of these investments going towards technology infrastructure. This quarter, we returned capital to shareholders through stock buybacks of $13.6 billion and dividend payments of $2.5 billion. Now, let's turn to our outlook section. I would like to clarify a few factors that will impact our business performance in the second half of 2025, as well as our updated expectations for full-year capital expenditures. First, regarding revenue, we are pleased with the growth momentum demonstrated by our current business overall. At current spot exchange rates, our revenue for the third quarter may receive a positive boost from exchange rates. However, exchange rate volatility may still affect the extent of the impact of exchange rates on revenue in the third quarter. As for our various business segments, in Google Services, advertising revenue in the second half of 2025 will be influenced by the following factors: First, the comparative effect brought about by our strong performance in the financial services vertical market during 2024; Second, the significant increase in advertising spending related to the U.S. elections in the second half of 2024 will negatively impact year-over-year comparisons, especially on YouTube. In the Cloud business, as I mentioned earlier, demand for our products remains strong, as evidenced by continuous revenue growth and a backlog of $106 billion in cloud business orders. Although we have been working to increase capacity and have accelerated the pace of server deployments, we still expect that demand and supply will remain tight until 2026. Turning to investments, given the strong demand for our cloud products and services, we now expect capital expenditures for 2025 to be approximately $85 billion, up from our previous estimate of $75 billion. Our updated expectations reflect: Additional investments in servers, changes in server delivery timing, and accelerated data center construction progress, primarily to meet the needs of cloud customers. Looking ahead to 2026, due to continued growth in customer demand and business expansion opportunities across the company, we expect capital expenditures to further increase. We will provide more details about the 2026 capital expenditure outlook in future earnings call meetings. Regarding expenses: First, as I mentioned in the previous quarter's earnings call, the significant increase in capital expenditures over the past few years will continue to put pressure on our profit and loss statement (P&L), primarily reflected in higher depreciation costs. In the second quarter, depreciation increased by $1.3 billion year-over-year, reaching $5 billion, with a growth rate of 35%. Given the recent increase in capital expenditure investments, we expect the growth rate of depreciation to further accelerate in the third quarter Second, as we mentioned earlier, we expect to see some growth in employee numbers in key investment areas in 2025. In the third quarter, we anticipate a quarter-over-quarter increase in total new hires, partly due to the recruitment of recent graduates. Third, the expenses in the third quarter will also reflect the costs associated with the upcoming launch of the new generation Pixel product series in August. To summarize, as you heard from Sundar and Philipp, we are pleased with the company's growth momentum and excited about our pace of innovation. Our full-stack approach—integrating AI infrastructure, AI research, and AI products and platforms—enables us to launch new products and services across the company. We have made significant progress in AI, as evidenced by the increase in the cumulative number of tokens we have processed. Revenue from our search business is growing steadily, with features like AI Overviews, AI Mode, and Google Lens providing users with new ways to access information. The annual revenue run rate for our cloud business has exceeded $50 billion, and we are continuously investing to meet customer demands while also improving profitability. YouTube is also expanding its addressable market, such as with Shorts, which now has an average daily view count exceeding 20 billion. We are very excited about the value our products and services bring to global customers and partners. Now I will hand the call back to the host, and next, I, Sundar, and Philipp will answer your questions. Q&A Session Host: The Q&A session is now starting. (Operator instructions) Your first question comes from Eric Sheridan of Goldman Sachs, your line is open. Eric Sheridan: Thank you very much for taking my question. I would like to ask Sundar and Philipp each a question. Sundar, as you think about the evolution of products and platforms, how do you view the impacts brought by changes in consumer behavior? How should investors understand this issue from the perspective of traffic and monetization? Because there are many long-standing dynamics, such as the relationship between clicks and click monetization, and as we look ahead to the next three to five years, these dynamics may change significantly. Philipp, regarding the development of YouTube, you just mentioned some aspects related to subscription revenue I am very curious about how you view the combination ratio between advertising revenue and subscription revenue? With the continuous expansion of subscription services, what key experiences or insights do you have in this area? Thank you. Sundar Pichai (CEO): Thank you, Eric. Thank you for your question. I do believe that from everything we see right now, people are excited about artificial intelligence, and they are widely adopting it across our various products. For me, seeing multimodal applications and how people are changing their behavior by naturally incorporating images into their interactions with Google, whether through Google Lens or Circle to Search, indicates that these features have been well accepted as part of their daily experience. This is just an early sign, but it suggests that people will adopt these new technologies well at these key moments. I understand your question may be inquiring about clicks and click monetization, and perhaps Philipp can elaborate on that part. But overall, as we continue to build new natural user experiences, we are very clear on how to continue training and optimizing monetization to integrate well with natural experiences and meet all the conditions required by those experiences. So for some emerging platforms, like the Gemini app, we will focus on creating excellent natural usage experiences in the near term, but just as we have done in AI Overviews and AI Mode, over time we will also be able to introduce very good commercialization experiences on these new platforms. And we believe that users will gradually adapt to these changes as they have in the past. Perhaps Philipp can add a bit more? Philipp? Philipp Schindler (Google Chief Business Officer): Okay, regarding the relationship between YouTube subscriptions and advertising that you mentioned, we can say this: We really like our advertising business, and we also really like our subscription business. YouTube's subscription service is becoming increasingly important, and we will definitely continue to focus on this in the long term. Our YouTube subscription products achieved strong growth this quarter, which includes — to clarify — YouTube TV, YouTube Music, and Premium. I think a common theme of our subscription services is: providing more choices for viewers. We also have a very deep understanding of the monetization side: for example, in which scenarios can we achieve higher monetization through advertising? And in which areas can we achieve higher monetization through subscriptions? Therefore, I believe we will actively continue to promote this "dual-track strategy" in the future, with advertising and subscription developing in parallel. Analyst: Thank you. Host: Our next question comes from JP Morgan's Doug Gnus, your line is now open. Analyst Doug Gnus: Thank you very much for answering my question. I have a question for both Sundar and Philipp. Sundar, could you talk about your current views on the ability to acquire computing resources? Although you have spent an additional $10 billion on capital expenditures this year, you mentioned that we are still in a tight supply environment, so I would like you to explain how these two factors are balanced? Then Philipp, regarding search growth, could you discuss the composition of the 12% growth in search revenue, specifically in terms of paid clicks and pricing growth? As we look to the future, how should we view the relationship between traffic growth and monetization trends? Thank you. Sundar Pichai (CEO): Thank you, Doug, and thank you for your question. Regarding capital expenditures, clearly, we are seeing strong growth momentum across our entire business portfolio, especially in Cloud. You are right that it is indeed a tight supply environment right now, and we are also increasing investments to expand capacity, but there is clearly a time lag in this — that is to say, our new investments now will gradually take effect over the next few years. So, these two things (increased investment and tight supply) can both be true and are not contradictory. But we are already planning ahead and continuing to invest, and overall, seeing our growth momentum in Cloud is exciting. I believe this is mainly due to the comprehensiveness of our AI product portfolio; our products cover a wide range, providing customers with GPU and TPU-based model services, which are greatly driving demand growth, so we are working hard to meet this demand through investment. Philipp Schindler (CBO): Regarding the paid clicks issue you mentioned, let me be very clear: We have mentioned before that our operational goal is to create quality results for users and provide attractive returns on investment (ROI) for advertisers. We do not manage the business with specific targets for "number of paid clicks" or "cost per click (CPC)." In fact, some of the product or policy changes we have made may sometimes reduce paid click volume while improving monetization efficiency. You can see in our submitted 10-Q filing that paid clicks have increased by 4% year-over-year this year, but these metrics are influenced by many factors each quarter, such as: advertiser spending, product changes, policy adjustments, user engagement, and so on. Therefore, it is very important not to draw overly broad conclusions based solely on these metrics when interpreting paid clicks and CPC. Doug Anmuth: Thank you both. Host: Our next question comes from Brian Nowak of Morgan Stanley, your line is now open. Brian Nowak: Thank you for answering my question, I have two questions. First, Sundar, there has been a lot of discussion about agentic search for commercial activities and broadly deployable agents. Perhaps you could talk from a technical perspective about what the main technical challenges are that the engineering team responsible for developing these new intelligent agent capabilities believes need to be overcome to launch a scalable commercial query agent? That’s the first question. Second, I remember you previously updated us on the statistics regarding the internal efficiency improvements we have achieved through generative AI (GenAI). Is there any latest progress in this area? Additionally, what friction points still need to be overcome for generative AI internal tools, and do you have any experiences or learnings to share? Thank you. Sundar Pichai (CEO): Thank you, Brian. Let me start by addressing the first question about intelligent agent capabilities. Overall, we have indeed invested the most resources in this direction when building the 2.5 series models, especially the Pro version. We have made exciting progress so far, including some models that we have not fully released yet. The main issue we need to address is that you are obviously chaining a series of events, so reliable execution is necessary, but latency can accumulate, costs can accumulate, and it must be done reliably in a way that users can accept, all of which must be coordinated. We are continuously making progress in these areas, and all parts must work together. The good news is that we have made solid progress and believe we are at the forefront in this field. Looking back over the past 12 months, you will find that we have been continuously improving the efficiency of our models to achieve the same capabilities at a lower cost. Therefore, the trajectory of future development will truly unlock the experiential potential of intelligent agents. We see the potential of this technology to realize these intelligent agents, but currently, they are still somewhat slow, costly, and sometimes perform weakly, but we are making progress in these areas I expect 2026 to be a year where intelligent agent experiences are more widely used, which will be an exciting opportunity. The second question, when you mentioned "internal efficiency sources," I guess you are referring to how we use these technologies internally. Regarding intelligent agents, we have already started to roll out intelligent agent coding workflows for our internal software engineers. Excitingly, in the past few months, especially in recent weeks, our internal software engineers have begun to use intelligent agent workflows extensively. This is a great example; a few months ago, there were many friction points with such workflows, but we are overcoming these issues, and employees are starting to use these tools in coding and other areas within the company. This is exciting progress, and I anticipate this will be an active area where we will also launch similar experiences for users. Stay tuned. Brian Nowak: Thank you, Sundar. That's fantastic. Host: Our next question comes from Michael Nathanson of MoffettNathanson, your line is now open. Michael Nathanson: Thank you, Sundar. I have two questions. At the IO conference, you announced the collaboration with Warby Parker to develop glasses. So I want to know how you view the importance of the new device cycle for further scaling AI, do you foresee a world where a smartphone is no longer at the core of our consumer experience? That's the first question. Secondly, what are the differences between the AI model used in Google Search and the Gemini standalone application? I want to know if there are differences in the types of users or usage patterns for both. Thank you. Sundar Pichai (CEO): Thank you, Michael. Regarding the first question, I believe any IO will drive new experiences, including hardware experiences. I think AI will particularly drive long-held expectations for glasses and other form factors. AI will spark a wave of new innovation. We are very excited about our investment in glasses, finding that the experience has significantly improved compared to the previous generation. I believe this will be an exciting emerging category, but I still expect smartphones to be at the core of the experience for the next two to three years. So I still think smartphones will continue to be at the core of the consumer experience, but we are also excited about emerging categories. Regarding the second question, between the AI model and the Gemini standalone application, overall, there are some use cases where both can provide a great experience, but there are also very specific use cases. I think when queries are information-oriented, users want to rely on authoritative information, Moreover, the AI model excels when they need the full capabilities of AI. You can use it there; it has a supportive foundation, and the Gemini model deeply utilizes search as a tool, so everything is based on the search experience. User feedback on this has been very positive. In the standalone Gemini application, users can engage in long conversations or chat to pass the time. In the Gemini app, you will see that early users sometimes enter a healing-like experience. These are all emerging experiences for users. I am also pleased that we have these two entry points to innovate in both areas simultaneously. Of course, some areas will be served by both applications. Over time, I believe we can make the user experience more seamless. Analyst: Thank you, Sundar. Host: Our next question comes from Mark Shmulik at Bernstein; your line is now open. Mark Shmulik: Thank you for taking my question. Sundar, there have been reports almost daily about the AI talent war and high-profile talent poaching; what is your take on this issue? How do you think Google is doing in attracting and retaining key AI talent? Additionally, how should we view the human resource costs related to AI compared to the significant capital investment in building AI-related infrastructure? Thank you. Sundar Pichai (CEO): Mark, regarding the first question, I think we have gone through similar phases before. Clearly, we have been deeply investing in talent, especially AI talent, for many years; we have been doing this for over a decade. I believe we have a very rich and deep talent pool. From my experience, top talent values: they want to be at the forefront driving progress, so the sense of mission and the advanced nature of the technology they work on are very important to them. Additionally, access to computing resources and the opportunity to work with peers—collaborating with the best in the industry—are all very important. Combining these factors, using these resources to make an impact. I believe we are very competitive in these areas. At this stage, we continue to focus on talent retention and the onboarding of new talent, and both are performing well. I know individual cases can make headlines, but looking deeply at the data, we are doing well and will continue to invest in talent and computing resources to ensure we are prepared for future opportunities. I will pass the microphone to Anat. Anat Ashkenazi (Senior Vice President and Chief Financial Officer): Yes, regarding how to integrate these costs into the overall cost structure, I mentioned earlier that we have a complete technology stack advantage, including research teams, which is one of our most important resources We ensure appropriate investments to attract the smartest talents in the industry, allowing them to drive innovation and serve customers at Google. This is reflected in the operating expenses across our organization. At the same time, we also strive to offset the costs brought by business growth and investments, ensuring reasonable allocation of resources. Sundar just mentioned the internal use of AI tools, which is another way we are enhancing business efficiency by optimizing operations through these tools. Additionally, we continue to manage the company with high standards of discipline and execution, driving business efficiency. Host: Our next question comes from Ross Sandler at Barclays, your line is now open. Ross Sandler: Great. I have two questions. The first is about search click-through rate (CTR) as a monetization driver. Over the past decade, you have done an excellent job in improving ad relevance and search click-through rates. Looking ahead, with the emergence of AI overviews and different forms of AI search, the number of ad displays on the search results page (SERP) has decreased. How do you view your ability to drive click-through rates in the future? The second question is, it seems you are now collaborating with OpenAI on certain cloud infrastructure work. Can you share how this partnership might expand in the future? Thank you. Philipp Schindler (Senior Vice President and Chief Business Officer at Google): I'll answer the first question. If I understand correctly, you are specifically referring to AI overviews. Sundar just mentioned them. AI overviews continuously enhance user satisfaction and increase the number of search users, They are scaling very smoothly, currently covering over 2 billion users across more than 100 countries, and we are very pleased with this development. As for monetization, we have discussed before that we see the monetization rate remaining roughly stable, which provides us with a solid foundation to innovate and launch more innovative next-generation ad formats. That is our current view. Sundar Pichai (CEO): Regarding the second question about collaboration with OpenAI, we are very pleased to work with them on Google Cloud. Google Cloud is an open platform, and we have extensive experience supporting excellent companies, startups, AI applications, etc., so we look forward to collaborating with them in the cloud and are eager to increase investment and expand this partnership. Host: Our next question comes from Mark Mahaney at Evercore, your line is now open. Mark Mahaney: Okay, may I ask two questions? First, Philipp, could you describe your view on the advertising environment for the second half of the year? Is it as certain or uncertain as last year for the second half of this year? The performance looks quite strong; do you have any unusual concerns for the second half? Then, Sundar, I would like to ask you again about the dual interface strategy for search. You clearly have some internal metrics showing that this is the optimal strategy. But some believe that a unified search interface, serving users by identifying the intent behind queries (whether it's purely informational or commercial), could give you a significant advantage in the market. Can you talk about the metrics you see and why you believe the dual interface solution is the best? Thank you. Philipp Schindler (Senior Vice President and Chief Business Officer of Google): I'll go first. We mentioned that the advertising business performed strongly in the second quarter. I can give you some insights from various verticals, The growth in search and other businesses in the second quarter covered all verticals. We noted that the biggest contributors were retail and financial services, mainly due to the strong performance in the insurance sector. Healthcare was also a significant contributor to growth. We have just entered the third quarter a few weeks ago, so it's too early to comment on what will happen in the second half. Sundar Pichai (CEO): Mark, regarding the second question, I believe these two interfaces cover the full breadth and depth of what humans might do. So there are good reasons for both interfaces to exist now. Of course, you are right that search is more information-oriented, while we see the Gemini application more as your assistant, a more personalized, proactive, and powerful assistant that serves various aspects of your life. You can imagine tasks like coding, creating long videos, etc., that the Gemini application does better now. Over time, just like the evolution we've experienced before, we will better understand user intent and abstract some complexity away from users. There was a time when people would search separately for text, images, videos, etc., and we made all of this seamless through unified search. We have the capability to integrate experiences to the extent that users find it appropriate and take on the heavy lifting for users. But I think, being in the early stages of an emerging paradigm, we want to ensure we meet current user expectations and provide a better experience over time. That's our current way of thinking. Analyst: Okay, thank you. Host: Our next question comes from Ken Gawrelski at Wells Fargo; your line is now open Ken Gawrelski: Thank you very much. If possible, I would like to ask two questions. The first is about the cloud business. I hope you can clarify your outlook for the second half of the year. Because last quarter you mentioned some supply constraints that are expected to ease by the end of 2025, but in the second quarter, you accelerated growth, and now you say the supply constraints will continue to ease until 2026. Please help clarify this a bit, considering the strong performance in the second quarter, your outlook for the cloud business in the second half of the year. The second is a bigger question about "agentic experience." Will it democratize the web like search did twenty years ago, allowing long-tail users to be discovered, or will it lead to more concentration with only a few winners in vertical industries? I hope you can share your thoughts on this issue. Thank you. Anat Ashkenazi (Senior Vice President and Chief Financial Officer): Okay, regarding your first question about the outlook for the cloud business in the second half of the year and the capacity growth I mentioned earlier. Clearly, we are working to bring more capacity online, meaning data centers and servers are being put into use, and we expect to see more growth in the second half of this year. But in reality, we are increasing capacity every quarter, as you can see, the growth rates for this quarter and the last quarter are quite good. As Sundar mentioned earlier, this investment is not a "switch" that can be turned on; it takes time to complete. So what you are seeing now is the result of investments we made some time ago, which are now starting to translate into new capacity coming online, but more of it will be reflected in the second half of the year. I also want to say that when considering cloud growth, it should not be viewed linearly, as the growth rates between quarters can be affected by the timing of capacity delivery, which can lead to fluctuations in performance between quarters. Sundar Pichai (Chief Executive Officer): Regarding "agentic experience," I think someone asked about the technological progress earlier. Clearly, the value propositions of all parties involved are key, and I believe this will be an important part of unlocking the future. As we advance in agentic experience, the experience for users will clearly improve. More savvy participants will leverage these experiences to help them grow and seize this opportunity. So I think this is an opportunity for some participants. You are right, just like the early internet, this will broaden access, increase use cases, and so on. These factors are all present. But I feel this is not just a technical issue; we also need to address the business model issues involving all parties, which is also an important part of future evolution. Ken Gawrelski: Thank you. Host: Our last question comes from Justin Post at BAML. Your line is now open Justin Post: Thank you. A few questions for Sundar. First, it seems that the subscription business is performing very well, especially with Gemini 2.5 receiving a lot of positive reviews. How is the progress of the Gemini subscription business? I know this is a key area of focus for the company. Do you have any measures in place to accelerate the growth of Gemini consumer subscriptions in Google One? Secondly, regarding the strategic adjustment of capital expenditures, the increase is significant, mainly due to cloud demand. Can you talk about the return on investment for cloud capital expenditures and what gives you confidence that this investment will yield good returns? Thank you. Sundar Pichai (CEO): Okay, the first question about subscriptions, Google One has always been a very attractive value proposition, primarily based on storage space. But now with our AI initiatives, including the Pro and Ultra versions, especially the 2.5 series models, these have indeed accelerated user conversion. This has been a very healthy quarter. So we are very excited about future opportunities. You will find that during this period, we have been able to drive growth in this area based on AI products. This is a field we are very optimistic about, especially since the launch of 2.5 Pro last quarter, we have seen significant progress. Regarding capital expenditures, particularly in the cloud segment, I believe we are investing because we are delivering a lot of value through our cloud products. It is important to understand that as the Google Cloud install base grows, customer satisfaction is very high, and churn rates are very low, we are more efficient in investing in the growth of this business. You can see this from the year-over-year and quarter-over-quarter profit margin trajectories over the past few years. All of this enhances our confidence in our investments, believing that we can achieve healthy returns. Especially in the AI era, the value we create for customers is also significantly increasing. So I believe these factors will help us succeed. Host: Thank you, everyone. The Q&A session for today is now concluded. I will now hand the meeting over to Jim Friedland for closing remarks. Jim Friedland (Senior Director of Investor Relations): Thank you all for participating today. We look forward to speaking with you again in the Q3 2025 conference call. Thank you, and have a good evening. Host: Thank you, everyone. The conference call is now concluded. Thank you for participating. You may now disconnect