Meta: What drives the high Ads growth in Q3? (Minutes of 2Q24)

The following is the summary of the second quarter financial report conference call of $Meta Platforms(META.US) in 2024. For financial report analysis, please refer to "Mag 7 Thunderous Thunder, Can Meta withstand the "clear stream"?" here.

I. Review of Core Financial Information:

II. Detailed Content of the Financial Report Conference Call

2.1. Key Points from Executive Statements:

1) Business Progress

① User Growth

Over 3.2 billion people use Meta's applications every day, with outstanding performance in the U.S.; WhatsApp's monthly active users in the U.S. have exceeded 100 million; Facebook, Instagram, and Threads have achieved significant growth globally; On Facebook, the engagement of the 18-29 age group has significantly increased, reflecting the success of strategies targeting this group.

② Significance of AI for the application family and core business

a. Improvement in recommendation systems: AI continues to enhance the content recommendation quality of Facebook and Instagram, driving user engagement. The full-screen video player and unified video recommendation service launched this quarter integrate Reels, long videos, and live content, further enhancing the application effectiveness of the AI system.

b. Development of advertising services: AI will gradually take over the generation and personalized customization of advertising creatives, with advertisers only needing to provide business goals and budgets in the future, and AI will handle the rest. This will significantly improve the precision and efficiency of ad placements.

③ New AI experiences and opportunities

a. Meta AI Assistant: After the extensive rollout of the Meta AI Assistant, it is rapidly improving in intelligence and functionality, expected to become the most widely used AI assistant.

b. AI Studio: The launch of AI Studio allows users to create their own AI agents within applications, helping creators interact with the community and supporting the development of commercial AI applications. AI Studio provides users with a richer interactive experience, especially bringing new opportunities for creators and businesses.

c. Llama base model: The Llama 3.1 model has been released, including open-source models and leading small and medium-sized models that outperform closed models in cost and performance. The open-source strategy will drive industry innovation and enhance the prosperity of the entire ecosystem④ The Significance of AI for the Metaverse

a. Metaverse Hardware: AI has driven the development of Metaverse products, with demand for Ray-Ban Meta glasses exceeding expectations, and the Quest 3 headsets not only performing well in the gaming field but also gradually becoming a general computing platform

b. Metaverse Ecosystem: Horizon's usage in virtual reality, mobile, and desktop continues to grow, expected to become an important part of the Metaverse ecosystem. With further integration of AI and Metaverse products, more innovative opportunities and use cases will be brought forth.

⑤ Future Outlook

The current strong business performance lays the foundation for deep investments in the future. Meta plans to continue investing in foundational technologies and product experiences to achieve long-term returns. The Meta team is confident in the future development and will continue to create value for the community and investors.

2) Financial Performance

① Overall Performance

a. Total Revenue: Total revenue for the second quarter was $39.1 billion, a year-on-year increase of 22%, or 23% on a fixed exchange rate basis

b. Total Expenses: Total expenses were $24.2 billion, a year-on-year increase of 7%

c. Cost of Revenue: Cost of revenue increased by 23%, mainly driven by infrastructure and Reality Labs inventory costs

d. Operating Expenses: Research and development expenses increased by 13%, mainly due to employee-related costs and infrastructure costs; marketing and sales expenses decreased by 14%, mainly due to lower restructuring and employee-related costs; administrative expenses decreased by 12%, mainly due to lower legal-related costs

e. Operating Income: Operating income was $14.8 billion, with an operating profit margin of 38%

f. Net Income: Net income was $13.5 billion, with earnings per share of $5.16

② Family of Apps Performance

a. Total Revenue: Total revenue for the family of apps was $38.7 billion, a year-on-year increase of 22%

b. Advertising Revenue: Advertising revenue for the family of apps was $38.3 billion, a year-on-year increase of 22%, mainly driven by growth in online commerce, gaming and entertainment, and media industries

c. Other Revenue: Other revenue for the family of apps was $389 million, a year-on-year increase of 73%, mainly driven by the growth in commercial messaging revenue from the WhatsApp Business Platform

d. Operating Income: Operating income for the family of apps was $19.3 billion, with an operating profit margin of 50%

③ Reality Labs Performance

a. Total Revenue: Revenue for Reality Labs was $3.53 billion, a year-on-year increase of 28%, mainly driven by sales of Quest headsets

b. Total Expenses: Expenses for Reality Labs were $4.8 billion, a year-on-year increase of 21%, mainly due to increases in employee-related costs and inventory costs

c. Operating Loss: Operating loss for Reality Labs was $4.5 billion

④ Capital Expenditures and Cash Flow

a. Capital Expenditure: Capital expenditure in the second quarter was $8.5 billion, mainly used for servers, data centers, and network infrastructure

b. Free Cash Flow: Free cash flow was $10.9 billion

c. Buybacks and Dividends: $6.3 billion of Class A common stock was repurchased this quarter, and $1.3 billion in dividends was paid to shareholders

d. Cash and Debt: Total cash and securities at the end of the quarter were $58.1 billion, with debt at $18.4 billion

⑤ Business Outlook

a. Revenue Forecast: Total revenue for the third quarter of 2024 is expected to be between $38.5 billion and $41 billion

b. Expense Forecast: Total expenses for the year are estimated to be between $96 billion and $99 billion, with a significant increase expected in the operating loss for Reality Labs

c. Capital Expenditure Forecast: Total capital expenditure for 2024 is projected to be between $37 billion and $40 billion, with a significant increase expected in 2025 to support AI research and product development

2.2 Analyst Q&A

Q: About new user and advertiser use cases supported by generative AI. Can you elaborate on aspects that interest you and explain their potential impact on the business in 2025 and 2026?

A: I believe that the greatest results in 2025 and 2026 will come from the impact of AI on existing products. AI is enhancing recommendation systems, helping users find better content, and improving the advertising experience. These products already have scale, and the application of AI will further enhance the experience and business outcomes. For new projects like Meta AI and AI Studio, they will increase user engagement in the short term and bring other benefits, but it will take a few years to achieve large-scale monetization. This is similar to what we have seen with Reels - the expansion and monetization of products take time.

Those who focus on our business in the long term can predict early signs of success several years in advance. That's why I mentioned early signs of Meta AI in my comments. We just launched Meta AI last quarter, and now after several months, I believe we are on track to become the most widely used AI assistant by the end of the year. While this is just the beginning, early signals indicate that we are on the right path. I believe a fundamental feature of AI is that it will eventually impact almost all of our products. It can not only improve existing products but also give rise to many new products. That's why tech company CEOs frequently discuss AI on earnings conference calls, as it will change many different things over various time frames.

Q: You have many capital expenditure priorities, including building new infrastructure and computing capabilities for next-generation models. Can you elaborate on your capital expenditure philosophy and how you ensure healthy returns for investors?

A: We categorize AI investments into core AI and generative AI. In core AI, we take an ROI-based approach and see strong returns as increased engagement and improved ad performance translate into revenue growth. Therefore, continuing to invest in this area is reasonable. In generative AI, we are still in the early stages and do not expect them to be a major source of revenue in 2024However, we expect that over time, generative AI will open up new revenue opportunities, especially in enhancing core advertising business, driving commercial information delivery, and increasing engagement with Meta AI.

In addition, we maintain flexibility when building AI infrastructure to ensure that it can be used for generative AI training and can be adjusted for inference, ranking, and recommendations. We adopt a phased approach to developing data centers, allowing us to quickly increase capacity when needed while controlling future expenses. Although capital expenditures are expected to increase significantly by 2025, we have a robust framework to address these investment opportunities and maintain flexibility.

Q: Regarding revenue guidance and prospects, what can you share about the overall digital advertising market situation?

A: We see that global advertising demand remains healthy, and we are enhancing ad performance (such as ad objectives, ranking, placements, etc.) to improve effectiveness. These improvements are expected to continue to drive ad spending in the third quarter. However, due to facing a high base from strong growth of Chinese advertisers last year and growth in Reels displays, year-over-year growth in the third quarter may slow down. Additionally, the headwinds from exchange rates in the third quarter are also expected to slightly increase.

Q: How do you view the rapid growth of platform elements like WhatsApp and Threads, and their potential between engagement growth and overall monetization?

A: The growth trend of WhatsApp in the United States is very important as the U.S. accounts for a significant portion of our revenue. In the past, WhatsApp was a leading messaging app in many countries globally but not in the U.S. Now, as more U.S. users discover that WhatsApp offers the best experience for cross-platform and group communication, this will bring a huge tailwind to our business growth in the U.S.

If the U.S. becomes a significant market for WhatsApp, it will be very beneficial for our overall business. Personally, it is gratifying to see more people around me using WhatsApp, but more importantly, its impact on the business.

Regarding Threads, I believe it is another successful example. Threads is the fastest-growing app, reaching 100 million users, which reminds us that even with such a rapid start, the growth path from 100 million to 1 billion users still takes many years. This is a characteristic of our product development - we first expand the user base and then gradually monetize it.

For Threads, the most exciting part is that such growth opportunities are very rare. We have been building this company for 20 years, and having another opportunity to create an app with the potential to reach 1 billion users is not common. If we can execute well, Threads could become another important part of our product portfolio. Although we are close to 200 million users, there is still a lot of work to be done to make it a significant part of the business. This opportunity is very rare, so we are very excited about it and satisfied with the team's performance in this regard.

At the same time, looking at recent sources of growth, video will continue to be the main driver of growth in the second half of the year. On Instagram, we expect Reels to continue driving growth, and on Facebook, we expect overall video viewing time to increase, especially the proportion of short videos, creating more growth opportunitiesIn addition, we also expect the overall application community to continue expanding.

Q: Regarding infrastructure and capital expenditures, you mentioned that the construction is not only for Llama 3 and 4, but may also extend to the 7th generation. Given that you have done so much construction in advance, how does this affect the trend of capital expenditures over the years?

A: First of all, we have not shared specific long-term capital expenditure trajectory. Part of the reason is that infrastructure is currently an extremely dynamic planning area, and we are continuing to study the scope of the generative AI roadmap. Obviously, we expect to significantly increase investment in AI infrastructure next year and will provide further guidance in due course. When constructing these capital expenditures, we consider how to build flexibly to deploy to core AI and generative AI use cases as needed. At the same time, we ensure that investments in core AI can bring satisfactory returns that can be measured immediately. We are also satisfied with the investment opportunities in generative AI, so we have maintained flexibility in building these infrastructures to meet future needs.

Q: Please talk more about the outlook for the third quarter, are there any other more specific drivers leading to the strong performance expected?

A: Regarding the revenue outlook for the third quarter. We have seen robust global advertising demand on the platform, and we continue to improve ad performance, which stems from efforts over multiple quarters and will continue to accumulate value for the platform. Revenue in the second quarter grew by 22%, showing strong performance in all regions and verticals, especially among smaller advertisers. We expect this trend to continue into the third quarter.

Q: Comments on the growth of young adult users in the United States, especially on Facebook and Instagram. What are these users specifically doing on Facebook and Instagram? Can you quantify the growth in usage?

A: Building products for young adults has always been one of Facebook's core priorities, and these efforts have translated into increased engagement among this group. Over the past few quarters, app usage by young adults in the United States and Canada has seen healthy growth. We see that "Groups" and "Marketplace" are particularly popular among young adults. In the United States and Canada, young adults are driving growth in "Groups" posts, and they are also active users of "Marketplace," benefiting from product improvements and strong demand for second-hand goods in the United States.

Q: Regarding "Marketplace," it may be larger than eBay, but with lower monetization. What are your thoughts on future monetization? Does it include improving the quality of "Marketplace"?

A: We are excited about the strong performance of "Marketplace" among young adults. More broadly, "Marketplace" is part of our overall business strategy, focusing on creating the best shopping experience on the platform. While "Marketplace" is consumer-facing, a larger part of the business strategy is to make it easier for businesses to promote products on our platform, attract buyers to find and purchase relevant items. Additionally, we are very pleased with the investment in "Shop Ads." "Shop Ads" revenue is growing at a strong year-over-year rate, not only enhancing the incremental performance of advertisers but also integrating well with other products (such as Advantage+ Shopping), driving overall effectiveness

Q: What is your view on the future growth of the employee count? Considering that capital expenditure will increase significantly, will the number of employees grow moderately or significantly?

A: Regarding the employee count, we maintain strict discipline in the allocation of new employees to ensure their focus on the company's core priorities while addressing past recruitment shortages. As we further address this issue this year, I anticipate that the number of on-site employees by the end of 2024 will be significantly higher than the end of 2023. We have not provided expectations for employee growth in 2025 as the budgeting process has not yet begun. However, future recruitment is expected to be primarily focused on priority areas, and we will continue to manage employees in a very disciplined manner.

Q: Llama 3.1's AI assistant has been applied in different versions and seems to be getting closer to becoming a full search engine that can handle almost all queries, except for commercial queries. Are there plans to open it up to a wider network or link it to third-party markets for commercial searches? Regarding Ray-Ban, can you discuss the opportunities for deepening cooperation with EssilorLuxottica? What are your most interested areas?

A: I am very excited about the performance of Llama 3.1. The team has progressed from Llama's first version to last year's Llama 2, and now to the current Llama 3.1, which is essentially competitive and even leading in some aspects compared to other top closed models. Meta AI has used a version of Llama 3.1 and other services to build a coherent product. While the initial usage trends are promising, we still need to add more features such as commercial queries and specific functionalities in vertical domains to make it an ideal AI assistant. This is a long-term roadmap expected to be gradually realized over the next few years and become a widely used service. We will continue to work on Llama. Llama is the core engine of Meta AI, it is open-source, and I am very excited about our progress in these two aspects.

In terms of smart glasses, EssilorLuxottica is an excellent partner. We have launched the second generation of Ray-Ban Meta glasses, which have performed better than expected, with faster growth than the first generation, aligning well with the AI revolution and new features. We hope to continue working with EssilorLuxottica to build the next generation of glasses and deepen AI products. Although still in the early stages compared to becoming truly leading consumer electronics products, the performance has been very positive.

Q: Regarding the product vision of Llama 3, is it possible to offer it to other companies, such as customer service or call centers, or other vertical domains? What are your insights on utilizing open source and Llama 3.1 to provide greater services to enterprises?

A: Llama is a foundational model that can be applied to various different products, whether it's Meta AI, AI Studio, business agents, or assistants in Ray-Ban glasses, all built on Llama. Similarly, developers can leverage Llama to build a variety of different applicationsOpen source is very valuable to us because we want to ensure that we have leading infrastructure to support the consumer and business experiences we are building. This infrastructure is not just isolated software fragments, but an ecosystem that brings together the different abilities of many developers, such as model customization, fine-tuning, and inference optimization.

If we keep Llama for internal use, its value is far less than opening it up to developers to enhance our product ecosystem. Therefore, we do not intend to build a cloud and sell it directly, but to collaborate with AWS, Databricks, NVIDIA, Microsoft's Azure, Google Cloud, and others to ensure that developers can use Llama anywhere.

In terms of enterprise applications, we are particularly focused on the business agent segment. In the future, almost every enterprise will have at least one business agent to help handle all transactions from sales to customer support. AI has great potential in this area, especially in countries with high labor costs. When companies can quickly integrate information and build agents, it will be a huge advantage. We will combine Llama with the customization work of the business team, along with other investments in the ecosystem, to make Llama better. This applies not only to the business field but also spans all the different product areas we are building. So there is a lot of exciting potential.

Q: In your interview with Jensen, you mentioned "if expansion stops one day, you will have 5 years of product work ahead of you." Besides agents or AI assistants, what other areas in the AI field are you focusing on, especially in the roadmap for the next 5 years?

A: That's an interesting question. While assuming there are no new base models, the industry will still bring a lot of product innovations, but at the same time, we expect new base models to emerge in the future that will unlock new capabilities. We are planning product roadmaps around these models.

Our roadmap includes Llama 4, Llama 5, and beyond, focusing on achieving broad practicality in Meta AI, allowing enterprises, creators, and individuals to build the AI agents they need. In addition, real-time, multimodal smart glasses will become more useful over time. There is a certain lag between when technology becomes available and when it is fully explored in products, but this is a very exciting field with a lot of innovation for a long time.

Q: Regarding capacity issues, how do you model the entire capital expenditure roadmap, including training, inference, and future needs?

A: We are building a significant amount of capacity, which is driving a substantial increase in our capital expenditure by 2025. We have not yet shared plans for the long-term outlook. When considering new data center projects, we plan how to use them throughout their lifecycle, first considering the capacity needed to train future generations of Llama and designing the architecture accordingly. At the same time, we also plan how to use them for other core businesses in the later stages of their lifecycle, including generative AI inference, advertising, and content ranking and recommendations

Therefore, we plan across a wide range of potential use cases throughout the lifecycle of each data center, considering needs ranging from generative AI training to inference support, core advertising, content ranking, and more. At the same time, we evaluate the types of servers required to support different use cases. This planning approach takes into account a long-term perspective to ensure that data centers can meet future needs, while also considering multiple decision points throughout the data center lifecycle to ensure we have enough flexibility when ordering servers and allocating usage, enabling us to make the best decisions in the future.

Risk Disclosure and Disclaimer for this article: Dolphin Research Disclaimer and General Disclosure