Google stated that it expects this year's capital expenditures to reach an "astonishing" 75 billion dollars, with around 16 billion to 18 billion dollars in the first quarter. The growth of AI search is even stronger, as younger users embrace AI in their searches. There is confidence in AI agents, and some features of Project Astra will be integrated into the Gemini application this year.
On February 4th, Tuesday, after the close of Post-Market Trading on the East Coast, Alphabet announced its financial data for the fourth quarter. Among them, despite a significant increase in spending, the cloud business, which directly benefits from the application of AI technology, grew less than Wall Street expected.

Key highlights from Google's parent company Alphabet's earnings call summary:
1. Despite disappointing performance from Google Cloud in the last quarter, Google is still increasing its investment in AI. The company announced that it expects capital expenditures to reach an astounding $75 billion this year, with approximately $16 billion to $18 billion in the first quarter. The projected total investment level may vary by quarter, primarily due to delivery and construction schedules.
2. Google Cloud is not invincible. For several recent quarters, Google Cloud has been a highlight of Alphabet's performance. However, this quarter's sales fell short of expectations, and the company's market share still lags behind Amazon and Microsoft. Google seems optimistic about this business, as Chief Financial Officer Anat Ashkenazi anticipates that there will be "growth in employee numbers in key areas like Cloud" this year, especially noticeable after years of layoffs.
3. The growth rate of GCP far exceeds that of overall cloud services, with two factors contributing to the slowdown in growth. First, Google had a very strong AdAi deployment quarter in Q4 2023. Second, there was indeed very strong demand for AI products in Q4 2024, ending the year with demand exceeding available capacity. Therefore, Google is in a tight supply-demand situation and is working diligently to increase supply.
4. The usage of Google's large model Gemini reached 4.4 million in six months. Younger users are embracing AI in searches. Google stated that the growth of AI searches is particularly strong across all user demographics, including younger users. In this AI era, search performance is excellent. Google believes it has an opportunity to drive further growth by making searches more convenient, allowing people to interact more easily and pose follow-up questions.
5. Podcasts have become a major advantage for YouTube. Alphabet's Chief Business Officer Philipp Schindler stated that the advertising for the U.S. elections has fueled revenue growth on YouTube. Schindler reported that the combined spending by the American Democratic and Republican parties for the 2024 election is nearly double that of 2020.
6. Google is confident in AI agents. Pichai stated that as technology evolves, AI agents will serve as a new way for Google to assist users. He mentioned that some features of Project Astra (an agent that processes visual inputs using a smartphone camera) will be integrated into the Gemini application this year.
Here is the full transcript of the conference call (translated by AI):
Host: Welcome everyone! Thank you for joining Alphabet's Q4 2024 and fiscal year earnings conference call. All participants are currently in mute mode. There will be a Q&A session after the speakers finish.
Sundar Pichai, Chief Executive Officer: Hello everyone. We had a strong performance in Q4, thanks to our leadership in AI and our unique full-stack approach. We have made significant progress in computational power, model capability, and efficiency improvement. We rapidly rolled out product enhancements and gained tremendous momentum in consumer and developer adoption.
We are pushing towards the next frontier in areas from AI agents, reasoning, and deep research to cutting-edge video and quantum computing. The company is in a good rhythm and cadence, building, testing, and launching products faster than ever. This is translating into product usage, revenue growth, and outcomes. In search, the AI overview has now launched in over 100 countries, continuing to drive higher satisfaction and search usage.
Meanwhile, the 'Circle to Search' feature is now available on over 0.2 billion Android devices. Regarding Cloud Computing Service and YouTube, we indicated at the beginning of 2024 that we expected combined annual revenue from Cloud Computing Service and YouTube to exceed 100 billion dollars by year-end. We achieved this goal, finishing with 110 billion dollars in annual revenue. We are ready for continued growth. So today, I will provide an update on our AI progress and how it improves our core consumer products. Then I will briefly touch upon Cloud Computing Service, YouTube, platforms and devices, and Wayfair.
Let's start with AI. Last quarter, I outlined three differentiated full-stack approaches to AI innovation: our leading AI infrastructure, our world-class research (including models and tools), and our products and platforms that scale these innovations to people. First, AI infrastructure. Our complex global network of Cloud Computing Service regions and Datacenters provides a powerful foundation directly to us and our customers, driving revenue.
We have a unique advantage because we developed every component of our technology stack, including Hardware, compilers, models, and products. This approach allows us to enhance efficiency across various levels from training and reasoning to productivity improvement. In 2024, we broke ground on 11 new Cloud Computing Service regions and Datacenter campuses in places like South Carolina, Indiana, Missouri, and around the world.
We also announced seven new submarine cable project plans to strengthen global connectivity. Our leading infrastructure is also one of the most efficient in the world. The computing power of Google's Datacenter has increased nearly fourfold compared to five years ago, significantly enhancing calculation capabilities per unit of electricity. These efficiencies, combined with the scalability, cost, and performance we offer, are reasons more and more organizations choose the Google Cloud Platform.
In fact, today, the computing power used by cloud customers for training and inference has grown more than eight times compared to 18 months ago. We will continue to invest in our cloud business to ensure we can meet the growing demand from customers.
Second, we have world-class research, including models. In December, we launched Gemini 2.0, which is our most powerful AI model to date, built for the era of agents. We released an experimental version of Gemini 2.0 Flash, which is our low-latency, performance-enhanced flagship model. Flash has already been launched in the Gemini applications, and tomorrow we will fully roll out 2.0 Flash to developers and customers, along with other model updates. Stay tuned.
At the end of last year, we also launched our experimental Gemini 2.0 Flash thinking model. The scaling progress of the thinking models has been very rapid, with extremely positive evaluations so far. We are developing better thinking models and look forward to sharing them with the developer community soon. The advancements in Gemini 2.0’s multimodal and native tool use enable us to build new agents, bringing us closer to realizing the vision of a universal assistant.
An early example is Deep Dive. It leverages agent capabilities to explore complex topics for you and provides key insights along with sources. It was launched in December in the premium version of Gemini and is being promoted to Android users worldwide. We have seen our consumer Gemini application gaining significant product momentum, which first debuted on iOS last November.
We have also opened access to some trusted testers of research prototypes, including Project Mariner, which can understand and infer information on browser screens to complete tasks, and Project Astra. We expect to introduce the capabilities of these two projects into the Gemini application later this year. We are also excited about our progress in video and image generation models. VO2 is our state-of-the-art video generation model, and Imagen 3 is our highest quality text-to-image model.
These generative media models, along with Gemini, rank at the top of the industry and have received the highest ratings in industry benchmarks. That’s why today over 4.4 million developers are using our Gemini models, a number that has doubled compared to six months ago. We continue to make breakthroughs in quantum computing research. At the end of last year, we announced Willow, our new state-of-the-art quantum computing chip that can reduce errors by several orders of magnitude as we use more qubits. Willow is an important step in our journey toward building practical quantum computers with real applications. This technology is promising, which is why this breakthrough has generated real excitement.
Third, our products and platforms bring AI to billions of people worldwide. We have seven products and platforms with over 2 billion users, all using Gemini. This includes search, where Gemini is supporting our AI overview. People are using search with AI overviews more, and as time goes on, as users realize they can ask new types of questions, usage is also increasing. This behavior is particularly evident among younger users, who highly appreciate the speed and efficiency of this new format.
We are also pleased to see how 'Circle to Search' drives additional search usage and opens up more types of questions. This feature is also popular among younger users. Those who have previously tried 'Circle to Search' now use it to initiate over 10% of searches. As AI continues to expand the range of questions people can ask, 2025 will become one of the biggest years for search innovation in history. Now let me turn to the key highlights of this quarter's Cloud Computing Service, YouTube, platforms and devices, and Waymo.
First, Google Cloud. Our AI-driven Cloud Computing Service has won us customers like Mercedes-Benz, Mercado Libre, and Servier. In 2024, the number of first-time commitments doubled compared to 2023. We have also deepened customer relationships. Last year, we signed several strategic deals worth over 1 billion dollars, doubling the number of transactions over 0.25 billion dollars compared to the previous year.
Our partners further accelerated our growth, with customers purchasing billions of dollars worth of solutions through our cloud marketplace. We continue to see strong growth in our broad portfolio of AI-driven Cloud Computing Services. This starts with our AI supercomputers, which offer leading performance and cost in both GPU and TPU. These advantages helped Citadel perform market modeling and training, and enabled Wayfair to modernize its platform, improving performance and scalability by nearly 25%.
In the fourth quarter, we saw strong adoption of our sixth-generation TPU—Trillium, which offers four times better training performance and three times the inference throughput compared to the previous generation. We also continue to maintain a strong partnership with NVIDIA. We recently delivered platforms based on their H200 to customers. Just last week, we were the first to announce customers running on the highly anticipated Blackwell platform.
The number of customers using our AI development platform Vertex AI has increased fivefold year-on-year, with brands like International and WPP building new applications and benefiting from our over 200 foundational models. Vertex usage increased twentyfold in 2024, especially due to the strong adoption of Gemini Flash, Gemini 2.0, Imagen 3, and the recent VEO. We have also seen strong growth in our AI-driven databases, data analytics, and Cybersecurity platforms.
Customers, including Radisson Hotels, are now using Gemini to search and analyze multimodal data from multiple clouds. Our AI-driven truck intelligence and security operation products help customers like Vodafone and AstraZeneca identify, protect, and defend against threats. Our growing portfolio of AI applications has also seen strong adoption from customers.
In the fourth quarter, we launched Google Agent Space, which helps businesses combine data with Google's quality search to create Gemini-driven agents and automate transactions for employees. Additionally, we have recently made our powerful Gemini AI capabilities available to all Google Workspace business and enterprise customers to help enhance their productivity.
Turning to YouTube. Nielsen's data shows that YouTube continues to rank first in streaming watch time in the USA, with our streaming share reaching an all-time high. On Election Day alone, over 45 million viewers in the USA watched election-related content on YouTube. Our early investment in podcasts is paying off. We have integrated podcasts into the core experience of YouTube, especially in combination with video.
According to the recent Edison report, we have now become the most frequently used Consumer service for podcasts in the USA. This success reflects our long-term investment strategy that moves from mobile to the living room, focusing on emerging trends. We have over 0.25 million creators in the USA and South Korea who have joined the YouTube Shopping affiliate program. At the end of last year, we expanded YouTube Shopping to three additional countries, allowing more creators to share their favorite products with fans and grow their business. Philipp will discuss YouTube's performance in more detail during the conference call.
Next up is platforms and devices. Google One performed exceptionally well, becoming one of our fastest-growing subscription products in terms of both subscribers and revenue growth. Last month, we announced the first paid Android 16 and new Android updates, including deeper integration of the upcoming Samsung Galaxy S25 series. We also recently announced Android XR, which is the first Android platform built for the twin star era.
Android XR is co-developed by Samsung and Qualcomm, aimed at powering the ecosystem for the next generation of extended reality devices (such as headsets and glasses). Lastly, a few words about Waymo, which made great strides last year, safely completing over 4 million passenger trips. It currently averages over 0.15 million trips per week and is still growing. Looking ahead, Waymo will expand its network and operational partnerships, entering new markets this year like Austin and Atlanta, and next year Miami.
In the upcoming weeks, Waymo One vehicles will arrive in Tokyo for their first international road trip. We are also developing the sixth-generation Waymo driver, which will significantly reduce hardware costs. Thanks to our global employees for another outstanding quarter. 2025 will be exciting, and we are ready. Philipp, I will pass the meeting to you.
Philipp Schindler, Senior Vice President and Chief Business Officer: Thank you, Sundar, hello everyone. I will briefly cover this quarter’s performance, then elaborate on our progress in search, advertising, YouTube, and partnerships, emphasizing the impact of AI on our business and customers. Google services revenue was $84 billion, up 10% year-on-year, mainly driven by an 11% growth in advertising revenue. Strong growth in search and YouTube advertising partially offset the year-on-year decline in network revenue.
In terms of vertical industry performance, search and other revenue grew by 13% year-on-year, with the financial services industry leading, followed by retail. YouTube advertising revenue grew by 14%, benefiting from strong advertising spending during the US elections, with total spending from both parties nearly double that of the 2020 elections.
In the fourth quarter, we saw sustained strong growth in search revenue. December brought many exciting updates, and we are quickly integrating our AI innovations into the consumer experience. We have begun testing twin star 2.0 in the AI overview and plan to roll it out more widely before the end of the year. In search, we have seen more people asking entirely new questions using voice, camera, or previously impossible methods, such as "Circle to Search."
We are providing these benefits to more consumers. Google is present in more than half of the journeys in which new brands, products, or retailers are discovered. By offering new ways to search, we are expanding business opportunities for our advertisers. Shoppers can now take pictures of products and quickly find product information, reviews, similar products, and where to purchase at favorable prices using Lens.
Lens is used for more than 20 billion visual search queries every month, most of which are new. The retail industry performed particularly strong this quarter, especially during Black Friday and Cyber Monday, where the advertising revenue for both days exceeded 1 billion dollars.
Interestingly, although the US holiday shopping season is the shortest since 2019, retail sales started in October, leading to a shopping season that lasted longer than expected. People shop on Google more than 1 billion times each day. Last quarter, we launched a redesigned Google Shopping experience, rebuilt from the ground up with AI. In December this year, the daily active users of Google Shopping in the USA increased by approximately 13% compared to the same period in 2023. In terms of search, travel and share another interesting trend, we have seen spending expand to 'Travel Tuesday.'
This contributed to a 20% year-on-year revenue increase for travel advertisers on Cyber Monday and Travel Tuesday. Turning to advertising, we continue to invest in AI capabilities in media buying, creativity, and measurement. As mentioned earlier, we believe AI will fundamentally change every aspect of the marketing value chain. In the past quarter, we observed that our clients are increasingly focused on optimizing their use of AI.
For example, Petco uses DemandGen campaigns to find a new audience of pet parents on YouTube in terms of targeting, creative generation, and bidding. Their advertising spend ROI is 275% higher than their social benchmark, and the click-through rate is 74% higher.
In media buying, we made YouTube creator takeovers available broadly in the USA and will expand to more markets this year. Creators know their audiences best, and creator takeovers help businesses connect with consumers through authentic and relevant content. In terms of creativity, we introduced new control features in PMax and simplified reporting to help clients better understand and reinvest in their best-performing assets.
Using asset generation in PMax, the creative asset production at Event Ticket Center increased fivefold, saving time and effort. Compared to their period of using manual assets, they also increased conversion rates by 300%. Finally, in terms of measurement. Last week, we made Meridian—our marketing mix modeling—commonly available to clients, helping more businesses reinvest in the creative and media buying strategies they know are effective.
According to Nielsen's analysis of marketing mix models, on average, Google's AI-driven video campaigns on YouTube have an advertising spend ROI that is 17% higher than manual campaigns. Turning to YouTube, we've seen strong revenue growth, thanks to continued increases in watch time from ad-supported and premium experiences. Our focus remains on building a streaming platform that enables creators to thrive and unlock the potential of AI.
Expanding our state-of-the-art video generation model, we announced VO2, which creates extremely high-quality videos across a wide range of topics and styles. It's inspiring to see how people are experimenting with it. We will roll it out to creators on YouTube over the coming months. We continue to invest in helping YouTube creators collaborate with brands. Now, all advertisers worldwide can promote YouTube creator videos and campaigns in Google Ads, and promote in all AI-driven campaign types, allowing creators to tag partners in their brand videos.
Sephora uses DemandGen's exclusive Short Video channel to enhance traffic and brand search for its holiday gift guide campaign, leveraging larger collaborations to find the best gifts. This has increased Sephora's holiday search volume by 82%. Short videos continue to rise and are closing the gap with long videos. In 2024, the monetization rate of short videos in the USA has increased by over 30 percentage points relative to long video views, and further progress is expected in 2025.
We are making it easier for advertisers to benefit from short videos, whether on big screens or small screens. We are particularly excited about its success on connected TVs, which now accounts for 15% of short video views in the USA. Louis Vuitton achieved its overall goals by combining ad formats on both long video and short video content.
Their short videos exceeded luxury goods benchmarks, with an average watch time 89% longer for the same length of video, while their long video content surpassed the benchmark by more than 15%, and garnered strong engagement among Gen Z and Millennials. Looking towards the living room, we continue to rank first in streaming watch time in the USA; according to Nielsen, this position has been maintained for nearly two years, and our streaming share has reached an all-time high.
Global audiences are watching over 1 billion hours of YouTube content daily through television in 2024. YouTube has invested years to adapt to changing consumer behaviors. The current search and living room viewing conditions directly reflect the efforts made over the years to build the right products and partnerships. Creators now prioritize a high-quality viewing experience that truly shines on TV screens, inspiring more audiences to watch.
In fact, the number of creators earning revenue primarily from television has increased by more than 30% year-on-year. We have also invested in podcasts, with popular shows like Club and Lex Fridman increasingly presented visually. YouTube creators and audiences are embracing this trend. In 2024, people will spend over 0.4 billion hours monthly just watching podcasts on living room devices. According to Edison research, YouTube is now the most popular podcast listening service in the USA.
As always, let me conclude with the strong momentum we see in partnerships, and the extensive services provided by Google are increasingly recognized. Sundar mentioned our deepening partnership with Samsung. Another expanding partnership is with Citi, which is leveraging Google Cloud to modernize its technology infrastructure to transform employee and customer experiences.
By using Google Cloud, it will improve its digital products, streamline employee workflows, and utilize high-performance computing to perform millions of calculations daily. This partnership also drives Citi's generative AI projects in customer service, document summarization, and search to reduce manual processing. That said, let me take a moment to thank the extraordinary commitment of Google employees worldwide, as well as the continued trust of our customers and partners in us. Anath, it’s your turn now.
Anat Ashkenazi, Chief Financial Officer: Thank you, Philip. We are pleased with the momentum shown across the business, as Alphabet's revenue reached $350 billion in 2024, an increase of 14% year-on-year, and a 15% growth at constant exchange rates compared to 2023. My comments will focus on year-over-year comparisons for the fourth quarter, unless otherwise stated.
I will start with the overall results of Alphabet, then cover our business unit results. I will conclude with comments and expectations for the first quarter and the full year of 2025. We achieved strong performance again in the fourth quarter, demonstrating robust momentum across the business. Consolidated revenue was $96.5 billion, a year-on-year increase of 12%, growing 12% on a constant exchange rate basis. Search remains the largest contributor to revenue growth, followed by Cloud Computing Service.
Total cost of goods sold was $40.6 billion, an increase of 8%. Technical costs were $14.8 billion, up 6%. We continue to see changes in revenue structure, with Google Search growing at a double-digit rate, while network revenue declined due to its higher technical cost rate. Other operating expenses amounted to $25.8 billion, up 9%, mainly due to increased content acquisition costs for YouTube, followed by increased depreciation resulting from our investments in technical infrastructure. The growth in content acquisition and depreciation was partially offset by a year-on-year decline in hardware costs, primarily due to the postponement of our 'Made by Google' product launches from Q4 2023 to Q3 2024.
In terms of total operating expenses, year-on-year comparisons reflect the $1.2 billion exit costs incurred from actions taken in Q4 2023 to optimize global office spaces. As previously disclosed, these costs are allocated between other operating expenses and operation expenses based on the number of relevant employees. Total operating expenses decreased by 1% to $24.9 billion. R&D investments grew by 8%, primarily due to increased compensation and depreciation costs, which were partially offset by the charges from office space optimization in Q4 2023.
Sales and marketing expenses decreased by 5%, primarily reflecting optimization expenses from last year, along with declines in compensation and advertising promotion costs due to the timing of Pixel product launches being moved up from Q4 to Q3. General and administrative expenses fell by 15%, reflecting changes in our charitable donations time and last year's optimization expenses. Operating income increased by 31% this quarter to $31 billion, with the operating profit margin rising by 32%, representing a 4.6 percentage point expansion in margin. Net income grew by 28% to $26.5 billion, with EPS rising by 31% to $2.15.
We achieved $24.8 billion in free cash flow in the fourth quarter and $72.8 billion in free cash flow for the entire year of 2024. We ended the quarter with $96 billion in cash and marketable securities. Turning to the business unit results. Google Services revenue grew by 10%, reaching $84.1 billion, reflecting strong momentum in Google Search and YouTube advertising. Google Search and other advertising revenue increased by 13% to $54 billion.
The strong performance of Search was again widely distributed across various verticals, with Financial Services leading due to strong performance in the Insurance Industry, followed by Retail. YouTube advertising revenue grew by 14% to $10.5 billion, driven by brand advertising and direct response ads. Network advertising revenue was $8 billion, down 4%.
In the fourth quarter, the year-on-year comparisons of all our advertising revenues were influenced by the growth in advertising revenues in Q4 2023, partially due to strong performance from retailers in Asia Pacific. Subscription platform and device revenue grew by 8% to $11.6 billion, primarily reflecting an increase in subscription revenue, partially offset by the shifting of our 'Made by Google' device launches from Q4 2023 to Q3.
We continue to achieve significant growth in subscription products, mainly due to an increase in paying subscribers for YouTube TV, YouTube Music Premium, and Google One. In terms of platforms, we saw a slight increase in growth rates for the Play Store, primarily due to strong growth in the number of buyers. Google Services operating income increased by 23% to $32.8 billion, with the operating profit margin rising from 35% to 39%, representing significant margin expansion. Turning to the Google Cloud business unit, this sector continued to achieve very strong results this quarter. Revenue grew by 30%, reaching $12 billion in Q4, reflecting growth in core GCP products, AI infrastructure, and generative AI solutions.
Once again, it is emphasized that GCP's growth rate is far higher than the overall Cloud Computing Service. The healthy growth of Google Workspace is primarily due to the increase in average revenue per seat. Google Cloud's operating revenue has grown to $2.1 billion, with an operating profit margin rising from 9.4% to 17.5%. We are satisfied with the Cloud team's efforts to provide valuable solutions to customers, drive revenue growth, and maintain a focus on improving the efficiency of the cloud business.
As for other investments, the revenue for the fourth quarter was $0.4 billion, with an operating loss of $1.2 billion. The year-on-year decline in revenue and increase in operating loss mainly reflect milestone payments made for one of the Other investment projects in the fourth quarter of 2023. Turning to Alphabet-level activities, the largest component of this line item is our investment in AI research and development activities that support the entire Alphabet. Just a reminder, Alphabet-level activities include almost all severance costs arising from layoffs as well as office space costs. The biggest factor in the year-on-year comparison for the fourth quarter of 2024 is the $1.2 billion expenses in the fourth quarter of 2023, which were nearly all related to office space optimization.
Regarding capital expenditures (CapEx), our capital expenditures reported for the fourth quarter were $14 billion, primarily reflecting our investment in technology infrastructure, with the largest portion being server investments, followed by Datacenter, to support the growth of Google services, Google Cloud, and Google DeepMind business. In the fourth quarter, we returned value to shareholders through $15 billion in Share Buyback and $2.4 billion in dividend payments.
Overall, we returned nearly $70 billion in value to shareholders in 2024. Looking ahead to 2025, a few factors that will impact our business performance in the first quarter and the entire year will be commented on. First, in terms of revenue, I will highlight two items that will significantly impact the company's revenue in the first quarter. The first is the impact of Exchange Rates.
At current spot Exchange Rates, we expect that the appreciation of the dollar against major currencies will create greater headwinds for our revenue in the first quarter of 2025 compared to the fourth quarter of 2024. The second is the impact of the leap year. We expect that revenue will face headwinds due to the fact that the first quarter of 2025 has one day less than the first quarter of 2024 due to the leap year. Regarding our business segments, Google's service advertising revenue in 2025 will be influenced by our strong performance in financial services throughout 2024. In the area of Cloud Computing Service, given that revenue is related to the timing of new capacity deployments, we may see fluctuations in the growth rate of cloud revenue depending on when the new capacity goes live in 2025.
Turning to investments, first is our expectations for capital expenditures for the full year of 2025. As mentioned in the third quarter conference call, as we scale our AI efforts, we expect to increase investments in capital expenditures on technology infrastructure, primarily in servers, followed by Datacenter and networks. We anticipate investing approximately $75 billion in capital expenditures in 2025, with about $16 billion to $18 billion in debt expected in the first quarter. The expected total investment level may vary by quarter, mainly due to delivery and construction timelines.
In terms of expenses, firstly, the increase in our capital expenditures investment over the past few years will exert pressure on the profit and loss statement (P&L), primarily manifesting as higher depreciation. In 2024, we saw a year-on-year increase in depreciation of 28% as we invested more in technology infrastructure assets. Given the increase in capital expenditure investments over the past few years, we expect the depreciation growth rate to accelerate in 2025. Secondly, we anticipate that the number of employees in key investment areas, such as AI and Cloud Computing Service, will see some growth in 2025.
As you just heard from Sundar, we are delivering products and solutions to customers at an incredible pace, building, testing, and launching products faster than ever before. As I mentioned in the third quarter conference call, we are doing this while also focusing on further improving the efficiency of our operations. Before we begin the Q&A, I would like to review the financial results for the year.
In 2024, revenue grew by 14%, totaling 43 billion USD, reaching 350 billion USD. Google Services and Google Cloud continued to achieve double-digit revenue growth, accompanied by margin expansion. The combined revenue from YouTube and Cloud Services reached an annual run rate of 110 billion USD by the end of the year. In 2024, we achieved a total revenue of 112 billion USD, an increase of 33% compared to 2023. We are pleased with the momentum we see in AI innovation and monetization.
We have been using AI in our advertising business to enhance performance for over a decade, and the Cloud Services are generating billions of dollars in annual revenue from AI infrastructure and generative AI solutions. We are also excited about the potential to bring new experiences to users, which will provide additional monetization opportunities. I look forward to sharing our progress throughout the year. Sundar, Philip, and I will now answer your questions.
Q&A Session
Host: Thank you. (Host prompts) The first question we received is from Brian Nowak of Morgan Stanley. Your line is now open.
Brian Nowak, Analyst: Thank you for taking my question. I have two questions, one for Sundar and one for Ananth. Sundar, regarding search, it seems there is still much room for improvement in the search space with the development of generative AI and agent technologies. Could you briefly outline your macro vision for the continued evolution of search products in the coming years, and how you see search products maintaining their forefront in helping users gather information while providing more engagement and monetization opportunities for users and advertisers?
The second question, Ananth, about 90 days ago, you mentioned further improving operating expenses (OpEx) efficiency and streamlining areas. Could you provide examples of what further efficiencies in operating expenses (OpEx) might look like besides the increase in depreciation and amortization (D&A) that we will see in 2025? Thank you.
Sundar Pichai, Chief Executive Officer: Thank you, Brian. Regarding search, clearly, this has been an ongoing journey. The AI overview is the next step; it is actively unfolding. As we pointed out, relevant metrics are performing well, and we are clearly leveraging that experience, continuously introducing better models, expanding the applicability of queries, and so on. But there is more to come.
I believe we will introduce AI in a more powerful multimodal way, like what we have already done through Lens and 'Circle to Search.' You can imagine future projects like 'Project Astra.' You can also envision the work we are doing in Gemini Deep Research, which greatly expands the types of use cases that search can apply to, which do not always provide instant answers but may take some time to respond. These are the areas we are exploring, and you will see us launch more new experiences for users in 2025.
Therefore, there is indeed a significant potential for development to be unlocked in the field of AI.
Anat Ashkenazi, Chief Financial Officer: Thank you. Regarding the question you mentioned about how we view future leverage and some comments I made in the last meeting, I certainly see opportunities to further increase productivity and efficiency, which is a priority area for us. We will do this to ensure that we continue to invest in areas like AI and Cloud Computing Service, where we see sustained growth potential.
I will continue to focus on the areas I mentioned earlier, which include technological infrastructure. The $75 billion capital expenditure I mentioned will mostly be used for our technological infrastructure, including Servers and Datacenters. Therefore, it is crucial to ensure that we make these investments in the most efficient way possible. Secondly, managing the growth in employee numbers will be important, and we will invest in growth areas like AI and Cloud Computing Service, while also balancing this growth across the company.
Optimizing the Real Estate footprint is another area I mentioned earlier. We continue to focus on this. Additionally, we are considering how to streamline the organizational structure. We previously talked about integrating similar fields, and Sundar mentioned bringing together some AI research teams so that we can operate at a faster pace, but we are also considering how to operate internally in the organization using our own AI tools to manage the Business, whether it is Sundar mentioning the use of AI to write code in the last meeting, or using AI tools to run some of our key processes.
We are considering all of these aspects. This will be an ongoing effort, not just a one-quarter thing. It will continue throughout the year, and we will continue to focus on it to support growth in other areas.
Host: The next question we received is from Doug Anmuth of JPMorgan. Your line is now open.
Doug Anmuth, Analyst: Thank you for taking my question. I have one question for Philip and another for Anat. First, Philip, could you elaborate on the expansion of advertising in the AI overview and anything else you might have learned in the fourth quarter? I'm curious, especially if you are targeting a higher proportion of commercial queries in advertising. And can it be said that your monetization level in existing searches is nearly equivalent?
Secondly, Anat, regarding the growth of Cloud Computing Service, it seems that there has been a slight slowdown from the third quarter to the fourth, but it sounds like you also hinted at capacity constraints in the fourth quarter. I would like to explore this issue further. Is this accurate? And can it be said that if there were more capacity, the revenue growth from Cloud Computing Service could be higher? Thank you.
Philipp Schindler, Senior Vice President and Chief Business Officer: Regarding your first question, first, the AI Overview, it's great to see it continuing to drive higher user satisfaction and search usage. As you know, we recently launched ads in the AI Overview on mobile devices in the USA, which expanded on our previous launch of ads above and below.
As I mentioned earlier, for the overall AI Overview, we actually see a monetization level comparable to existing search, which I believe provides us with a very strong foundation on which we can innovate further.
Anat Ashkenazi, Chief Financial Officer: Regarding the question about Cloud Computing Service, first, I'm pleased that we achieved $12 billion in revenue at the end of the quarter, with a year-on-year growth of 30%, which is quite impressive. As I mentioned in my prepared remarks, GCP's growth is far surpassing that of the overall Cloud Computing Service.
Regarding the two factors for the slowing growth. First, we had a very strong AdAi deployment quarter in Q4 2023. Secondly, as you mentioned, we did see very strong demand for our AI products in Q4 2024, finishing the year with demand that exceeds available capacity.
Therefore, we are in a tight supply and demand situation, and we are working very hard to add more capacity. As I mentioned, we are increasing our capital expenditure investments in 2024, continuing to expand in 2025, which will bring more capacity throughout the year. Thank you both.
Moderator: Your next question comes from Eric Sheridan of Goldman Sachs. Your line is now open.
Eric Sheridan, Analyst: Thank you for taking my question. I have just one question. Sundar, a little over two weeks ago, news out of China, I think investors have been asking a lot about the long-term cost curve of AI, as it transitions from infrastructure to application, or from training to inference, and perhaps even the custom silicon becoming more dominant throughout the theme.
I'd love to hear your thoughts on the news from a few weeks ago and what it might mean for Alphabet in the long term.
Sundar Pichai, Chief Executive Officer: Thank you, Eric. First, there are many views on DeepSeek. First of all, I think it is a very outstanding team. I believe they have done a very, very good job. For us, it is clear that over time, there will be cutting-edge model development, but you can really improve efficiency to better serve these models.
If you look at one of the shining areas of the Gemini model, it is in the cutting-edge fields of cost performance and latency. If you look at these three attributes, I think we are at the forefront of this field. I would say that our 2.0 Flash model and 2.0 Flash thinking model are among the most efficient models, including compared to DeepSeek's V3 and R1. I believe this is largely due to our investment in developing and optimizing the entire stack, our obsession with the cost per query, and all of this has served us well in handling the workload of serving billions of users, whether in our products or in Cloud Computing Service.
What I want to say is that if you look at the trends over the past three years, the ratio of spending on inference has been increasing compared to training, which is a good thing because obviously, inference is to support businesses with good ROI. Therefore, I think this trend is positive. I believe inference models, if any, will accelerate this trend, as it is clearly expanding in the inference dimension as well.
So, I think part of the reason we are so excited about AI opportunities is that we know we can drive extraordinary use cases because the cost of actually using it will continue to decline, which will make more use cases possible. That is the opportunity space. This is a huge opportunity, which is why you see us investing to meet this moment.
Host: Thank you. Your next question is from Michael Nathanson of MoffettNathanson. Your line is now open.
Michael Nathanson, Analyst: Thank you. I have two questions, one for Philipp and one for Anath. First, Philipp, my question is that we are starting to see more AI tools on E-Commerce websites, such as AI recommendation research on Google Shopping. Can you please talk about how these products and other AI tools are affecting shopping behaviour, and what impact this has on monetization?
Secondly, I would like to ask a question regarding long-term capital intensity. It seems there are some restrictions in the construction aspect, but regarding modeling future capital intensity, how should we think about it? And when considering if this is the right level of spending, what are you expecting?
Philipp Schindler, Senior Vice President and Chief Business Officer: That is a great question. We have been leveraging our advancements in AI to make it much easier to search for products on Google, which is obvious. In the fourth quarter, we actually launched a completely revamped Google Shopping experience, which we rebuilt from scratch using AI. People shop on Google more than 1 billion times a day. Last quarter, we launched this brand-new Google Shopping experience.
In December, we saw that the daily active users of Google Shopping in the USA increased by about 13% compared to the same period last year. So, this is a good development. Regarding the new Google Shopping experience, especially in response to your question, it can intelligently show users the most relevant products, helping to speed up and simplify your research process. You will receive an AI-generated briefing that includes key points to consider while searching, as well as products that may meet your needs.
So shoppers generally want low prices. The new page not only includes deal-finding tools like price comparison, price insights, and price tracking, but also a new dedicated personalized deals page where you can browse deals selected for you, all built on AI. Therefore, we believe this is a very interesting opportunity.
Anat Ashkenazi, Chief Financial Officer: Regarding capital expenditures, I think you might have mentioned two issues. One is capital intensity and the other is the return on invested capital. First of all, we are certainly looking forward, but we are doing so in a very responsible manner, with a very strict and even internal governance process, considering how to allocate these capacities and what we need externally to support customer demand, and this is also the case within the Google—Alphabet business.
As you just heard about comments on Cloud Computing Service, our demand indeed exceeds our available capacity. Therefore, we will work to address this issue and ensure we bring in more capacity. We have a very broad business, and we can reallocate capacity, whether it is through Google services or Google Cloud to support, as I said, whether it is Search, Google DeepMind, or Google Cloud customers, we can do this in a more efficient way.
We also examine every investment we make to ensure we do so in the most cost-efficient way to optimize our datacenters. As you know, our strategy primarily relies on our own design and construction of datacenters. Therefore, they lead the industry in terms of cost and large-scale electrical utility efficiency. We have our own custom TPUs, which are tailored for our own workloads.
Thus, they provide outstanding performance in terms of capital expenditure efficiency. Therefore, when we decide how to advance capital investments in the coming years, we will consider all of these factors.
Michael Nathanson, Analyst: Thank you.
Host: Your next question comes from Mark Shmulik of Bernstein. Your line is now open.
Mark Shmulik, Analyst: Yes, thank you for accepting my question. Sundar, I would like to follow up on Brian's earlier question. It seems there is suddenly very strong momentum in the field of AI consumer agents with your own Project Mariner efforts and recent releases from competitors, seemingly catching up to the old vision of Google Duplex.
As you look ahead a few years, where do you think consumer agents are headed, and what does that really mean for Google Search, beyond Lens? Is there enough room for both to thrive together, or will they eventually conflict? Thank you.
Sundar Pichai, Chief Executive Officer: First of all, I think we have definitely seen a lot of advancements in the underlying capabilities of these models. Gemini 2.0 is certainly built with the view of enabling more agent use cases. So, we have definitely seen internal progress, and I believe we will be able to provide our users with more agent experiences. Look, I actually think all of this is expanding the opportunity space.
Historically, we have always had informational use cases, but now you can act on your informational needs in a deeper way. When we talk about products like Google Assistant, that has always been our vision. Therefore, I believe the opportunity space is expanding. I think there is a lot of room here; this is far from a zero-sum game. I believe there are many new types of use cases that have room to thrive. For us, we clearly see additional use cases that we can begin to solve for Google Search users.
All early AI work has indicated that users will respond positively to these improvements. Therefore, I am optimistic about the future.
Host: Your next question is from Ross Sandler at Barclays. Your line is now open.
Ross Sandler, Analyst: Okay. I have one question about infrastructure and another about revenue expectations. First, if we look at the inference cost per 1 million tokens, and not at the API pricing for products like Gemini and GPT-4, but at the raw cost of generating inference tokens on your TPU architecture, do you think you are more cost-efficient in generating 1 million tokens compared to your peers in the Cloud Service space? Do you believe that as everything shifts towards inference, this is an advantage?
Secondly, Anant, you mentioned that the strong performance in financial services will become an issue in 2025, could you quantify that? Is it similar to the situation with outbound advertisers in the Asia-Pacific region that you mentioned earlier? Can you provide some numbers about this headwind? Thank you very much.
Sundar Pichai, Chief Executive Officer: Ross, starting from the DPU project, our V1 is essentially an inference chip. Therefore, we have always adopted an end-to-end stack approach so that we can achieve strong differentiation in cost, latency, and performance, especially in the frontier fields mentioned. I believe our full-stack approach and TPU efforts provide meaningful advantages. We plan to continue this, and you have already seen this.
I know you are asking about costs, but our pricing also reflects this differentiation. This is also why we are able to launch products like the Flash model, which has a very attractive value proposition, and this is driving developer growth. We have doubled the number of developers in about six months to 4.4 million. The usage of Vertex has grown 20 times in the past year. All of these are direct results of this strategy.
Anat Ashkenazi, Chief Financial Officer: Regarding the strong performance in the financial services category that I mentioned, this is mainly related to structural changes in the insurance sector. Therefore, it is more specifically reflected in the financial services field, particularly the insurance portion. We have seen this sustained performance, but it was a one-time growth, and we have seen this trend throughout the year.
I won't provide specific figures that we expect to see in 2025, but I am satisfied with the strong performance we see across industries, including the retail sector, where we ended this year with strong momentum. If anything, I want to emphasize that when considering this year, the Forex impact that I mentioned, and the revenue we missed in the first quarter due to one less day, are factors that should be noted.
Host: Your next question is from Justin Post of Bank of America. Your line is now open.
Justin Post, Analyst: Okay, thank you. I have two questions for Philip. First, you mentioned that the AI overview drove an increase in search usage. I just want to know if the overall search usage has accelerated with your greater integration of AI? I know competitors are quickly increasing their traffic in the AI field, but I want to know if you really have seen a true increase in the total amount of information collected.
Secondly, regarding YouTube, considering the shift from more professional content to user-generated content, how has this influenced your usage, and how do you view the impact of this shift on profit margins? Thank you.
Sundar Pichai, Chief Executive Officer: Regarding overall search usage, our indicators are healthy. We continue to see search usage increasing year over year. Of course, within this, the growth of the AI overview is even stronger, especially among all user groups, including younger users. Therefore, it has been well received. But overall, I think search is performing well in this AI era.
As I mentioned before, we have more innovations this year. I believe the products will evolve further. I think that as search becomes more convenient, allowing people to interact and ask follow-up questions more easily, we have the opportunity to drive further growth.
Philipp Schindler, Senior Vice President and Chief Business Officer: Regarding your question about YouTube, YouTube advertising performed very well overall in the fourth quarter, with brand advertising and direct response advertising driving growth. The advertising for the USA elections boosted brand revenue, and the spending we saw was almost double that of 2020. I mentioned this point. We also received strong contributions from the financial, retail, and technology sectors.
Therefore, overall, the operational metrics are strong, and watch time continues to grow robustly, particularly in key monetization areas such as short video and the living room. Regarding user-generated content (UGC) that you specifically mentioned, we have a very close relationship with creators, and we have always said that creators are the core of YouTube's success. They are the most important thing to us, and this strong creator relationship gives us great confidence.
Today, we have over 3 million channels that have joined the YouTube Partner Program, which is an incredible program. Therefore, we are very confident about our position and future development.
Host: The last question we received is from Ken Gawrelski of Wells Fargo. Your line is now open.
Ken Gawrelski, Analyst: Thank you very much. I would like to ask two questions. I want to focus on Gemini and consumer engagement. Sundar, there are reports that you have an ambitious goal to significantly increase the usage of Gemini by the end of 2025. Regarding this issue, I have two questions.
First, how should we think about the approach you will take to achieve this goal? Is it through more aggressive independent product marketing for Gemini, or by integrating it more closely into existing experiences, whether it's search, email, maps, etc.? Second, how should we think about the future monetization opportunities for Gemini?
At present, it is mainly a premium subscription service or a free service. Over time, do you see the potential for an advertising component? Please share any relevant information. Thank you.
Sundar Pichai, Chief Executive Officer: First, we have brought strong growth momentum to the Gemini application in the second half of 2024. Part of this is because we have made it more accessible. For example, we launched a dedicated application on iOS, which has received very positive reception and indeed gained significant traction. Therefore, driving natural growth through product launches.
We just launched our 2.0 series models last week. So 2.0 Flash, I think it is one of the most powerful models you can access for free. This has obviously also played a driving role. Thus, we have made rapid iterations there. We have made several key innovations there. Gemini Live I think has definitely been well-received by users, and the in-depth study of Gemini has been the same for advanced users.
Therefore, I think the combination of innovation and continuous product improvements that are making it better is driving a lot of usage, and we will launch more this year. We obviously have partnerships with Samsung. So there are other things that will contribute to this. In terms of monetization, of course, we are currently focused on the free tier and subscription services. But as you see at Google, we always want to prioritize user experience, and we do have some good ideas about the native advertising concept, but you will see us prioritizing user experience.
But we are indeed committed to making products that can serve billions of users, and advertising has always been an important aspect of this strategy. Therefore, as you see on YouTube, we will provide people with choices over time. But this year, I think you will see us focusing on the subscription direction.
Ken Gawrelski, Analyst: Thank you.
Host: Thank you. This concludes our Q&A session for today. I would like to turn the meeting back over to Jim Friedland to see if he has any further remarks.
Jim Friedland, Senior Director of Investor Relations: Thank you all for participating today. We look forward to speaking with everyone again in the Q1 2025 conference call. Thank you, and have a pleasant evening.
Host: Thank you all. Today's conference call has come to an end. Thank you for your participation. You may disconnect at any time.
Editor/ping