Google executives interpret Q 2 financial report: The most important task is to solidly promote Gemini 4
Alphabet reported second-quarter cloud business revenue that exceeded Wall Street expectations, but sales related to its search engine business were slightly lower than expected, which may exacerbate market concerns about its large investment in artificial intelligence. Alphabet’s second-quarter revenue reached $119.8 billion, a year-on-year increase of 24% After the financial report was released, Google CEO Sundar Pichai, Chief Commercial Officer Philipp Schindler and CFO Anat Ashkenazi held a conference call with analysts to answer questions about the business. Here's a summary of analysts' conference call answers: Morgan Stanley analyst Brian Nowak: I have two questions. First of all, Sundar, as time goes by, we see more generative artificial intelligence (GenAI) products and tools appearing on the market, and companies continue to increase their investment. Can you talk a little bit about where you see the return on investment (ROIC) for the entire generative AI space now compared to a year ago? Specifically, have your views changed now on the size of the opportunity and the timing of the return on investment? Sundar Pichai: In my opinion, we are still in the very early stages. I believe that generative AI will drive structural change with long-term impact in multiple areas. As far as our core information business is concerned, the current cutting-edge AI capabilities allow us to realize various possibilities, and I always believe that we still have a lot of work to do to truly transform these advanced capabilities into consumer-oriented product experiences. For example, you can imagine an end-to-end intelligent agent, so that AI can truly complete more and more valuable tasks for users. I think these represent huge development opportunities. If we can seize these opportunities, I believe this will also bring us a very impressive return on investment. The same is true for the enterprise user market. Our product demand performance also reflects this, and you can also observe it from our growth rate and other performance indicators. Moreover, when I communicated with CEOs and enterprise users, I found that most companies have actually just begun to explore the potential of AI technology, and there is still a long way to go before real value is released. In the past, we often discussed the development of cloud computing. At that time, enterprise workloads actually moved to the cloud, accounting for only a small portion of the total workload. Now, you can also think about a similar question: What proportion of workloads are currently truly AI-native or AI-empowered? In my opinion, this ratio is still very low, and it can be said that we are still in the early stages of the development of AI technology. From the perspective of return on invested capital, we adopt a full-stack strategy. Whether it is in the consumer market, the enterprise market, or from the perspective of the developer ecosystem, we have seen continued growth momentum. All in all, if there has been any change in our thoughts and ideas in the past year, it is that we are more optimistic about future development opportunities than a year ago, and we are more firmly convinced of the long-term value contained in these opportunities. Anat Ashkenazy: Regarding the issues you mentioned about future capital expenditures and how to deal with computing power supply constraints, the current environment we are in has not fundamentally changed, and supply is still the main factor limiting business development. We have actually mentioned this to you repeatedly for several consecutive quarters. Whether it is the demand from external users of Google Cloud or the demand for computing resources from various internal businesses of the company, they have actually maintained a very strong growth trend. Our capital investment principle has never changed: we will continue to increase investment as long as we believe that this investment can generate an attractive return for the company. As Sundar just said, we are confident in these current investment opportunities. In terms of capital planning, we will still maintain a long-term perspective. In other words, we will not only conduct overall planning from the perspective of business development needs in the next few years, but also focus on the next year and shorter-term needs, and continue to accelerate infrastructure construction to meet the growing market demand.
Over the past three years, we have significantly increased our infrastructure capacity, but market demand continues to grow faster than our expansion rate. Therefore, like the entire industry, we are still in an environment of tight computing power supply and resource constraints. At the same time, we are also actively improving our supply capabilities. Google adopts a full-stack layout strategy. This advantage allows us to continue to improve the operational efficiency and technical efficiency of technical infrastructure, thereby achieving higher computing power under existing resource conditions. Therefore, we will continue to invest as long as we see attractive investment opportunities. JP Morgan analyst Doug Anmuth: I also have two questions. First of all, Sundar, could you please talk about your confidence that the Gemini model will continue to maintain the world's leading level? This week Google just released a new model of the Gemini Flash series. However, compared to some other industry-leading large model laboratories, Google does not seem to release products as frequently as they do, nor does it have such a high market volume. So I'd also like to take this opportunity to hear your thoughts on outside concerns: Can Google continue to develop and maintain industry-leading large model capabilities? In addition, can you ask the management to talk to us about Google's layout plan in the field of AI programming? And how does management plan to further narrow the gap with competitors in this segment of the enterprise market? My second question is for Anat. Recently, Google has conducted equity and debt financing. Could you please tell us how the company currently views the optimal capital structure? At the same time, in the process of advancing financing decisions, how do management weigh "debt financing costs" and "equity financing costs"? Currently, the competition for global cutting-edge large models is in a stage of rapid evolution, with very fast technology iterations. At any point in time, the industry landscape may change. Google has always had a world-leading model, and we still maintain an industry-leading position in many key capabilities. Of course, there are also some areas that we believe still need further improvement, such as AI programming and agent coding. The team is currently concentrating resources to continue to optimize these capabilities. Take Gemini 3.6 Flash, released this week, as an example. The Flash series is our core and most widely used model product, and it is also one of the most in demand models at present. It achieves a very good balance between performance, cost, reliability and response latency, so it is widely welcomed by users. At present, we have widely used Gemini Flash in the entire product system, including network security, data analysis and other solutions. Taking the user service scenario as an example, enterprises not only need high-quality voice capabilities and real-time interaction capabilities, but also need models with real-time reasoning capabilities. For professional service organizations, they rely more on high-quality content summary, text generation and other capabilities. The Gemini Flash model has shown excellent performance in these application scenarios. In terms of agent coding, we are also continuing to iterate rapidly and will continue to launch new versions in the future. For example, compared with Gemini 3.5 Flash, the performance of the newly released Gemini 3.6 Flash in the DeepSuite benchmark test has improved by more than 10 percentage points; at the same time, in the process of achieving performance improvement, Token usage efficiency has also been further improved. At present, we have fully applied this model within Google, and are also testing its application effect in programming scenarios with a number of enterprise users. In terms of time, from the previous generation of products to today's latest version, we have made significant progress in only about six weeks. In the future, we will continue to maintain this rapid iteration rhythm.
As for the competition of global cutting-edge models mentioned in your question, we are full of confidence and will always be firmly committed. For the next generation of global leading models, we believe that a larger-scale basic model is needed as support. Currently, we have started pre-training for Gemini 4 and have set very high goals. I am very excited about the development progress of Gemini 4 that I have seen so far, and I believe that when it is officially released, it will bring surprises to users. In order to continue to maintain our leading position in the world in the next stage, we need to have a larger and more capable basic model like Gemini 4. Therefore, I think our most important task at the moment is to solidly advance this work and do our best. Regarding the capital structure and recent equity and debt financing issues mentioned in your question. When we plan future investments, we will look at the next year as well as assess our funding needs in the coming years from a longer-term perspective. First, we will evaluate how much cash flow generated from operating activities can support the scale of investment. As you can see from the results announced today, Google continues to maintain strong and healthy operating cash flow, which has always been the company's main source of funding. On this basis, we will comprehensively consider debt financing. We have significantly expanded our debt financing over the past 12 months. About a year ago, the company's debt balance was approximately US$16 billion; now it has increased to approximately US$100 billion, and a debt financing system covering multiple currencies and regional markets has been formed, with more diverse debt financing channels. In addition, while supporting the company's continued growth, we hope to maintain a strong and resilient balance sheet. This is also an important consideration for our recent entry into the equity financing market. For now, we have no plans to raise equity again, except for special arrangements. As everyone knows, our previous equity financing plan included part of the ATM (At-the-Market) issuance plan. In the future, the company will continue to use this mechanism to offset the share dilution caused by stock-based compensation and pay the tax costs related to stock-based compensation. Overall, in terms of capital allocation, we will mainly focus on three dimensions for overall planning and balance, including cash flow from operating activities, debt financing scale and equity financing arrangements. Our goal is to maintain a healthy and robust balance sheet while meeting ongoing investment needs. Goldman Sachs analyst Eric Sheridan: I have two questions about TPU (Tensor Processing Unit). Sundar, could you please talk about Google’s important experiences and insights as it continues to expand the scale of TPU deployment? Specifically, on the one hand, as the application scope of TPU continues to expand, how does management view the future market demand for TPU? On the other hand, in the next few years, when the demand for TPU from external customers and the demand for self-developed chips from Google's internal business are growing at the same time, how will the company allocate and balance resources between the two? I have a follow-up question. Anat, in the briefing you mentioned that TPU has had a positive impact on Google Cloud's business. Could you please give us more details? For example, what proportion of TPU-related business currently accounts for orders that have been signed but have not yet confirmed revenue in Google Cloud? In addition, how will TPU-related businesses gradually be converted into revenue in the next few years? What impact will it have on the company's profit margins? First, we are very pleased with the progress of our TPU product roadmap. Both the performance and the competitive advantages it brings have met our expectations, so we also use TPU extensively internally. As for the allocation of TPU resources, our first principle has never changed: first, we must ensure that there are enough TPUs to support Google's global leadership in the field of general artificial intelligence (AGI). This is the basis of all our work. At the same time, the market demand for computing power is very strong, so we also need to balance between internal demand and external customer demand.
For most Google Cloud customers, we currently mainly use TPUs and GPUs to run and provide Google's own AI model services, such as Vertex AI, Gemini Enterprise and other products, as well as various AI agent applications that are currently growing rapidly, all rely on these computing resources. As for those customers who want to use TPU directly as infrastructure, they want to rent TPU computing power to run their own models. We are expanding infrastructure deployment to meet this demand. For example, we are evaluating and promoting more solutions to deploy TPU to customers' data centers or to the data centers of other partners. This is the case with the projects we are currently working on with Blackstone. In this way, we hope to better balance the allocation of computing resources: on the one hand, we can ensure the research and development needs of the most advanced AI models as much as possible; on the other hand, we can also continue to support Google’s own AI model services for consumers and enterprise customers. Regarding the revenue recognition method for TPU system sales, you can understand it this way: when we sign the relevant agreements mentioned previously, these contracts will be included in Google Cloud's backlog. Most of the current $514 billion Google Cloud backlog comes from customer contracts for Google Cloud Platform (GCP), but it also includes orders related to TPU system sales. After signing the agreement, we will build inventory in advance to be able to deliver these TPU systems in the future. Because we need to produce and prepare in advance during the process of business expansion and scale upgrading, this process will be reflected in the cash flow from operating activities. And when we start delivering these TPU systems to customers, we typically start recognizing revenue from that stage. During the quarter, the revenue we recognized represented only a small portion of the overall contract value. In the future, we will continue to increase the delivery scale in 2026, and most of the revenue from this agreement is expected to be recognized in 2027. Barclays analyst Ross Sandler: I want to go back to the current “War of the Big Models” within the industry. Sundar, I would like to ask further about the speed of model release. We noticed that Google recently reached a third-party computing power cooperation agreement with SpaceX. What I want to know is, in addition to this initiative, what other measures has Google taken to speed up the iteration and release rhythm of the Gemini model, thereby further improving the update speed? You just mentioned the Gemini Flash series models, and their performance is very successful, which is consistent with your introduction. However, considering that the competition in the low-cost model market is already very fierce and there are many players, do you believe that the market positioning of Flash models is still the right direction? What are management’s views on the future development trends of the AI model market and how Google will participate in it? I want to talk about two points. First of all, we have been thinking about how to make the company's products cover the complete competitive landscape of AI models. For Google users, we hope to provide the best model selection at different price bands. That is to say, on the one hand, we attach great importance to building industry-leading top-level models; on the other hand, we also need to provide users with models with strong performance, lower cost, and suitable for large-scale applications. Therefore, you can see that we have launched different versions of large models such as Gemini Flash-Lite and Flash Pro. Our goal is to cover the entire AI model market, from the highest performance cutting-edge models to cost-effective models, to fully meet the needs of different users. Regarding model release speed, we will continue to speed up iteration. For example, we released Gemini 3.5 Flash at the Google Developer Conference, and then launched Gemini 3.6 Flash. In the future, you will see this series of models continue to be updated, and we will continue to improve in terms of agent reasoning capabilities.
At the same time, we are also investing a lot of resources to promote the research and development of Gemini 4. This is a very important project. Our goal is that with the release of Gemini 4, it can maintain a leading level in the competition of AI cutting-edge models at that time. Therefore, we are focusing on investing a lot of computing power and R&D efforts to promote the realization of this goal. In the process of building Gemini 4, we will also build a more powerful basic platform. On this basis, we will be able to launch subsequent versions more quickly in the future and continue iterative upgrades. Improving the model update speed and achieving a model release rhythm of about once a month is one of our important plans in the process of developing Gemini 4 major models. MoffettNathanson analyst Michael Nathanson: My first question to Sundar is again about the model competition issue that has been discussed on today's call. I would like to ask you to talk about this. In this new AI model competition, what do you think is Google’s real long-term competitive advantage? In other words, even if all companies are able to achieve similar levels of large model capabilities in the future, what strategic advantages do you think Google has that will help the company continue to maintain its growth momentum? What advantages do you think keep Google ahead of the competition? Anat, I also want to follow up on Eric's question about TPU. I would like to know more about what impact TPU may have on the company's profit margin in the future? Specifically, will the profit margin brought by the TPU business be higher than Google Cloud's current overall profit margin, thereby increasing the overall profit margin? Or is it that due to factors such as hardware investment and infrastructure costs, the TPU business will cause profit margins to be lower than the current Google Cloud business level? First, the levels of solutions we provide are very rich. One of the values of our "full stack layout" is that users choose Google not just to buy a model, but to get a complete solution. For example, in the field of network security or data analysis, customers deploy complete business solutions, of which the model is only a part. Taking network security as an example, customers can use products such as Chronicle, Wiz, and our upcoming CodeMender to use AI tools to discover security vulnerabilities, fix vulnerabilities, etc. Another example is the field of data analysis. In the past, enterprise data was often scattered in different systems and isolated from each other. Now, users can integrate these scattered data and build an intelligent analysis layer on top of the data through Gemini Enterprise. Overall, in these applications, the model is only one component of the overall solution. I think this is very important. Even in those scenarios that appear to be pure use of.
Over the past three years, we have significantly increased our infrastructure capacity, but market demand continues to grow faster than our expansion rate. Therefore, like the entire industry, we are still in an environment of tight computing power supply and resource constraints. At the same time, we are also actively improving our supply capabilities. Google adopts a full-stack layout strategy. This advantage allows us to continue to improve the operational efficiency and technical efficiency of technical infrastructure, thereby achieving higher computing power under existing resource conditions. Therefore, we will continue to invest as long as we see attractive investment opportunities. JP Morgan analyst Doug Anmuth: I also have two questions. First of all, Sundar, could you please talk about your confidence that the Gemini model will continue to maintain the world's leading level? This week Google just released a new model of the Gemini Flash series. However, compared to some other industry-leading large model laboratories, Google does not seem to release products as frequently as they do, nor does it have such a high market volume. So I'd also like to take this opportunity to hear your thoughts on outside concerns: Can Google continue to develop and maintain industry-leading large model capabilities? In addition, can you ask the management to talk to us about Google's layout plan in the field of AI programming? And how does management plan to further narrow the gap with competitors in this segment of the enterprise market? My second question is for Anat. Recently, Google has conducted equity and debt financing. Could you please tell us how the company currently views the optimal capital structure? At the same time, in the process of advancing financing decisions, how do management weigh "debt financing costs" and "equity financing costs"? Currently, the competition for global cutting-edge large models is in a stage of rapid evolution, with very fast technology iterations. At any point in time, the industry landscape may change. Google has always had a world-leading model, and we still maintain an industry-leading position in many key capabilities. Of course, there are also some areas that we believe still need further improvement, such as AI programming and agent coding. The team is currently concentrating resources to continue to optimize these capabilities. Take Gemini 3.6 Flash, released this week, as an example. The Flash series is our core and most widely used model product, and it is also one of the most in demand models at present. It achieves a very good balance between performance, cost, reliability and response latency, so it is widely welcomed by users. At present, we have widely used Gemini Flash in the entire product system, including network security, data analysis and other solutions. Taking the user service scenario as an example, enterprises not only need high-quality voice capabilities and real-time interaction capabilities, but also need models with real-time reasoning capabilities. For professional service organizations, they rely more on high-quality content summary, text generation and other capabilities. The Gemini Flash model has shown excellent performance in these application scenarios. In terms of agent coding, we are also continuing to iterate rapidly and will continue to launch new versions in the future. For example, compared with Gemini 3.5 Flash, the performance of the newly released Gemini 3.6 Flash in the DeepSuite benchmark test has improved by more than 10 percentage points; at the same time, in the process of achieving performance improvement, Token usage efficiency has also been further improved. At present, we have fully applied this model within Google, and are also testing its application effect in programming scenarios with a number of enterprise users. In terms of time, from the previous generation of products to today's latest version, we have made significant progress in only about six weeks. In the future, we will continue to maintain this rapid iteration rhythm.
As for the competition of global cutting-edge models mentioned in your question, we are full of confidence and will always be firmly committed. For the next generation of global leading models, we believe that a larger-scale basic model is needed as support. Currently, we have started pre-training for Gemini 4 and have set very high goals. I am very excited about the development progress of Gemini 4 that I have seen so far, and I believe that when it is officially released, it will bring surprises to users. In order to continue to maintain our leading position in the world in the next stage, we need to have a larger and more capable basic model like Gemini 4. Therefore, I think our most important task at the moment is to solidly advance this work and do our best. Regarding the capital structure and recent equity and debt financing issues mentioned in your question. When we plan future investments, we will look at the next year as well as assess our funding needs in the coming years from a longer-term perspective. First, we will evaluate how much cash flow generated from operating activities can support the scale of investment. As you can see from the results announced today, Google continues to maintain strong and healthy operating cash flow, which has always been the company's main source of funding. On this basis, we will comprehensively consider debt financing. We have significantly expanded our debt financing over the past 12 months. About a year ago, the company's debt balance was approximately US$16 billion; now it has increased to approximately US$100 billion, and a debt financing system covering multiple currencies and regional markets has been formed, with more diverse debt financing channels. In addition, while supporting the company's continued growth, we hope to maintain a strong and resilient balance sheet. This is also an important consideration for our recent entry into the equity financing market. For now, we have no plans to raise equity again, except for special arrangements. As everyone knows, our previous equity financing plan included part of the ATM (At-the-Market) issuance plan. In the future, the company will continue to use this mechanism to offset the share dilution caused by stock-based compensation and pay the tax costs related to stock-based compensation. Overall, in terms of capital allocation, we will mainly focus on three dimensions for overall planning and balance, including cash flow from operating activities, debt financing scale and equity financing arrangements. Our goal is to maintain a healthy and robust balance sheet while meeting ongoing investment needs. Goldman Sachs analyst Eric Sheridan: I have two questions about TPU (Tensor Processing Unit). Sundar, could you please talk about Google’s important experiences and insights as it continues to expand the scale of TPU deployment? Specifically, on the one hand, as the application scope of TPU continues to expand, how does management view the future market demand for TPU? On the other hand, in the next few years, when the demand for TPU from external customers and the demand for self-developed chips from Google's internal business are growing at the same time, how will the company allocate and balance resources between the two? I have a follow-up question. Anat, in the briefing you mentioned that TPU has had a positive impact on Google Cloud's business. Could you please give us more details? For example, what proportion of TPU-related business currently accounts for orders that have been signed but have not yet confirmed revenue in Google Cloud? In addition, how will TPU-related businesses gradually be converted into revenue in the next few years? What impact will it have on the company's profit margins? First, we are very pleased with the progress of our TPU product roadmap. Both the performance and the competitive advantages it brings have met our expectations, so we also use TPU extensively internally. As for the allocation of TPU resources, our first principle has never changed: first, we must ensure that there are enough TPUs to support Google's global leadership in the field of general artificial intelligence (AGI). This is the basis of all our work. At the same time, the market demand for computing power is very strong, so we also need to balance between internal demand and external customer demand.
For most Google Cloud customers, we currently mainly use TPUs and GPUs to run and provide Google's own AI model services, such as Vertex AI, Gemini Enterprise and other products, as well as various AI agent applications that are currently growing rapidly, all rely on these computing resources. As for those customers who want to use TPU directly as infrastructure, they want to rent TPU computing power to run their own models. We are expanding infrastructure deployment to meet this demand. For example, we are evaluating and promoting more solutions to deploy TPU to customers' data centers or to the data centers of other partners. This is the case with the projects we are currently working on with Blackstone. In this way, we hope to better balance the allocation of computing resources: on the one hand, we can ensure the research and development needs of the most advanced AI models as much as possible; on the other hand, we can also continue to support Google’s own AI model services for consumers and enterprise customers. Regarding the revenue recognition method for TPU system sales, you can understand it this way: when we sign the relevant agreements mentioned previously, these contracts will be included in Google Cloud's backlog. Most of the current $514 billion Google Cloud backlog comes from customer contracts for Google Cloud Platform (GCP), but it also includes orders related to TPU system sales. After signing the agreement, we will build inventory in advance to be able to deliver these TPU systems in the future. Because we need to produce and prepare in advance during the process of business expansion and scale upgrading, this process will be reflected in the cash flow from operating activities. And when we start delivering these TPU systems to customers, we typically start recognizing revenue from that stage. During the quarter, the revenue we recognized represented only a small portion of the overall contract value. In the future, we will continue to increase the delivery scale in 2026, and most of the revenue from this agreement is expected to be recognized in 2027. Barclays analyst Ross Sandler: I want to go back to the current “War of the Big Models” within the industry. Sundar, I would like to ask further about the speed of model release. We noticed that Google recently reached a third-party computing power cooperation agreement with SpaceX. What I want to know is, in addition to this initiative, what other measures has Google taken to speed up the iteration and release rhythm of the Gemini model, thereby further improving the update speed? You just mentioned the Gemini Flash series models, and their performance is very successful, which is consistent with your introduction. However, considering that the competition in the low-cost model market is already very fierce and there are many players, do you believe that the market positioning of Flash models is still the right direction? What are management’s views on the future development trends of the AI model market and how Google will participate in it? I want to talk about two points. First of all, we have been thinking about how to make the company's products cover the complete competitive landscape of AI models. For Google users, we hope to provide the best model selection at different price bands. That is to say, on the one hand, we attach great importance to building industry-leading top-level models; on the other hand, we also need to provide users with models with strong performance, lower cost, and suitable for large-scale applications. Therefore, you can see that we have launched different versions of large models such as Gemini Flash-Lite and Flash Pro. Our goal is to cover the entire AI model market, from the highest performance cutting-edge models to cost-effective models, to fully meet the needs of different users. Regarding model release speed, we will continue to speed up iteration. For example, we released Gemini 3.5 Flash at the Google Developer Conference, and then launched Gemini 3.6 Flash. In the future, you will see this series of models continue to be updated, and we will continue to improve in terms of agent reasoning capabilities.
At the same time, we are also investing a lot of resources to promote the research and development of Gemini 4. This is a very important project. Our goal is that with the release of Gemini 4, it can maintain a leading level in the competition of AI cutting-edge models at that time. Therefore, we are focusing on investing a lot of computing power and R&D efforts to promote the realization of this goal. In the process of building Gemini 4, we will also build a more powerful basic platform. On this basis, we will be able to launch subsequent versions more quickly in the future and continue iterative upgrades. Improving the model update speed and achieving a model release rhythm of about once a month is one of our important plans in the process of developing Gemini 4 major models. MoffettNathanson analyst Michael Nathanson: My first question to Sundar is again about the model competition issue that has been discussed on today's call. I would like to ask you to talk about this. In this new AI model competition, what do you think is Google’s real long-term competitive advantage? In other words, even if all companies are able to achieve similar levels of large model capabilities in the future, what strategic advantages do you think Google has that will help the company continue to maintain its growth momentum? What advantages do you think keep Google ahead of the competition? Anat, I also want to follow up on Eric's question about TPU. I would like to know more about what impact TPU may have on the company's profit margin in the future? Specifically, will the profit margin brought by the TPU business be higher than Google Cloud's current overall profit margin, thereby increasing the overall profit margin? Or is it that due to factors such as hardware investment and infrastructure costs, the TPU business will cause profit margins to be lower than the current Google Cloud business level? First, the levels of solutions we provide are very rich. One of the values of our "full stack layout" is that users choose Google not just to buy a model, but to get a complete solution. For example, in the field of network security or data analysis, customers deploy complete business solutions, of which the model is only a part. Taking network security as an example, customers can use products such as Chronicle, Wiz, and our upcoming CodeMender to use AI tools to discover security vulnerabilities, fix vulnerabilities, etc. Another example is the field of data analysis. In the past, enterprise data was often scattered in different systems and isolated from each other. Now, users can integrate these scattered data and build an intelligent analysis layer on top of the data through Gemini Enterprise. Overall, in these applications, the model is only one component of the overall solution. I think this is very important. Even in those scenarios that appear to be pure use of.