Meta: Advertising remains stable and strong, with continuous investment, striving to become the next generation computing platform (4Q23 conference call minutes)
Here are the key points from the Meta Platforms Q4 2023 conference call. For a review of the financial report, please refer to the article "Meta Platforms: China's Explosive Expansion and Zuckerberg's Generous Gift".
1. Management Report
Despite this being an efficiency-focused year, we have still achieved strong growth. User engagement on the platform has increased, and we have launched several new products, including Threads, Ray-Ban Meta Platforms smart glasses, and the MR device Quest 3.
Our next goal is to establish the most popular and advanced AI products and services. We strive to provide a world-class AI assistant for every user of our services. Every content creator will have an AI assistant to engage with the community, every merchant will have an AI assistant for interacting with shopping users, and every developer will have access to an open-source large-scale model for product development.
I also believe that everyone needs a new type of computing device to ensure smooth interaction with AI. This device should be able to see and hear what the user sees and hears, such as smart glasses.
(1) AI
I have come to realize that the next generation of services requires a comprehensive general intelligence. Previously, I thought that using AI to achieve social, commercial, and media tools mainly involved solving the challenges and problems of AI integration.
But now I understand that our large-scale models need inference, planning, encoding, memory, and other cognitive abilities to support the services we want to provide. We have been researching general intelligence for over 10 years, and Meta Platforms has extensive experience in integrating new technologies into our services. There are several key aspects to our vision:
1. Building a world-class computing infrastructure.
By the end of this year, we will have approximately 350,000 H100s, along with other GPUs, which is equivalent to the computing power of 600,000 H100s. In terms of computing power, we are in a favorable position.
The training volume of large-scale models is currently growing at a rate of 10 times per year. Our AI training cluster is just a part of our overall infrastructure, and the other computational demands are not growing as fast.
Having such a reserve of computing power may generate some controversy as it affects Capex. However, overall, I hope that Meta Platforms will continue to invest in infrastructure and build the most advanced clusters.
In addition, we are also constructing new data centers and designing custom chips for our business.2. Next is the open-source software infrastructure
Our long-term goal is to build an open-source general infrastructure while maintaining the uniqueness of our specific products. In the field of artificial intelligence, this general infrastructure includes our Llama model, including the currently training and progressing well Llama 3, as well as industry-standard tools we have developed, such as PyTorch.
This open-source approach has unleashed a lot of innovation in the industry, and we firmly believe in it. I know some people have questions about how we benefit from open-source, research results, and the openness of a large amount of computing resources. So I think it might be useful to explain the strategic benefits here.
In short, open-source improves the performance of our models. Since a lot of work is required to turn our models into products, and there will always be other open-source models available, we find that being an open-source leader is primarily advantageous and does not significantly diminish the differentiation of our products. More specifically, there are several strategic benefits.
(1) Open-source software is more secure and reliable, and it receives continuous feedback from the ecosystem, resulting in higher development efficiency. Both security and efficiency are important for AI.
(2) Open-source software is easier to become an industry standard. Once our stack is standardized, it will be easier to integrate new technological innovations into our products.
(3) Open-source is often more popular among developers. Developers are motivated to develop open systems that can be widely adopted, which also helps us recruit the best talent and maintain a leading position in the industry.
I want to emphasize again that because we have unique data and products, open-source Llama will not weaken our competitive advantage, which is why we are willing to make open-source ecology our long-term strategy.
3. Large models require continuous and ongoing development
We plan to advance the research of Llama 5, 6, and 7 in the coming years to develop a more comprehensive general intelligence.
4. Learning from unique data and feedback loops
FB and Instagram have billions of publicly shared photos and videos, which we estimate to be more than the amount commonly crawled elsewhere, and we also have a large amount of publicly available post and comment data.
However, we also know that in the early stages of training, it is crucial to establish the correct feedback loop so that hundreds of millions of people can participate in experiencing and interacting with our AI services. We can leverage this feedback to improve the effectiveness of Reels and other advertising (AI automatically recommends and distributes Reels content).
5. Rapid learning and testing of products
In this process, we aim to achieve the optimal product experience. We usually engage in organized learning and iteration, but there are also products that have gained popularity among users before they are fully tested and refined, such as Threads. I would also like to mention that Threads is still in the growth stage (130 million monthly active users), and I believe this product will be successful.(II) Metaverse
In addition to AI, our long-term vision also includes the Metaverse, and we will continue to invest in it. In the fourth quarter, Reality Labs' revenue exceeded 1 billion for the first time, and the Quest series performed strongly during the holiday season (both Quest 2 and 3 performed well, with the highest number of Quest app downloads on Christmas Day in the App Store). Our internally developed VR game, "The Wrath of Asgard 2," made a significant contribution.
Quest 3 has made a good start, and I hope it will continue to be the most popular MR device. This year, our focus is on developing the mobile version of Horizon and further improving the VR version.
Ray-Ban Meta Platforms smart glasses have shown strong user engagement, with better user stickiness and retention than the first generation glasses. Currently, there is a high demand, and our partner's production plan exceeds our original expectations.
One of the current product iteration plans is to first launch multimodal artificial intelligence and then introduce other truly exciting AI features later this year. I believe that users want a new hardware device to interact frequently with AI without having to take out their phones, press a button, and point it at what they want to see.
I think smart glasses will be a great example of how to combine the Metaverse and artificial intelligence.
(III) Advertising
We will continue to use AI to improve our product experience and advertising business. Reels and our discovery feature (recommendation algorithm) are still the main drivers of increasing user stickiness. At the same time, I believe that Messaging will be our next focus for generating revenue.
Currently, Reels is performing well on IG and FB, with users sharing Reels videos 3.5 billion times per day. Reels has now brought in net revenue for our applications. The biggest opportunity in the future is to integrate Reels and other types of videos using our recommendation system to help users discover better content on our platform.
Threads is steadily growing, with over 130 million monthly active users. I am optimistic about its growth and believe it can become a popular social application.
II. Key Financial Indicators (Full version can be found in the Earnings Report analysis)
Revenue of $40.1 billion, a growth of 25%, with a constant currency (cc) growth of 22%.
Total expenses of $23.7 billion, a YoY decrease of 8%.
Operating costs decreased by 8%, mainly due to a decrease in restructuring expenses, but there was also an increase in some infrastructure costs.
Research and development expenses increased by 8%, mainly due to increased employee costs in the app family and Reality Labs, as well as non-labor operating expenses in RL.
Sales expenses decreased by 29%, mainly due to a decrease in sales expenses and restructuring expenses YoY.
General and administrative expenses decreased by 26%, mainly due to a decrease in restructuring expenses YoY.
At the end of the fourth quarter, the number of employees was 67,300, a decrease of 29% compared to a year ago, but an increase of 2% compared to the third quarter. Our recruitment efforts have resumed.
Capex, including principal payments for financing leases, was 7.9 billion, driven by investments in servers, data centers, and network infrastructure.
FCF was $11.5 billion, reflecting a $1.6 billion deferred income tax payment from the previous quarters of 2023. Cash and marketable securities at the end of the year were $65.4 billion, with long-term debt of $18.4 billion.
In the fourth quarter, we repurchased $6.3 billion worth of stocks, bringing the total stock repurchase for the year to $20 billion. As of the end of the year, we had $30.9 billion remaining from the previous authorization, and today we announced an additional $50 billion authorization for stock repurchases.
III. Business Explanation and Guidance
Starting in the first quarter of 2024, we will no longer disclose FB's MAU, DAU, and the MAP of the Meta Platforms ecosystem. Instead, we will report the YoY changes in ad impressions and average prices in each region. DAP will continue to be disclosed.
The vertical sectors that contributed the most to YoY growth in advertising revenue were e-commerce, daily consumption, entertainment, media, and gaming. E-commerce and gaming benefited from strong customer acquisition demand from Chinese advertisers.
In 2023, revenue from Chinese advertisers accounted for 10% of our total revenue and contributed 5 percentage points to the overall revenue growth (excluding the impact of China, revenue grew by 20% YoY).
In terms of geographical distribution, advertising revenue growth was strongest in Europe and other parts of the world, with growth rates of 33% and 32% respectively, followed by the Asia-Pacific region at 23% and North America at 19%. Exchange rates had a slight impact on international revenue (estimated at 2-3 percentage points by Dolphin Research).
Ad impressions increased by 21%, with a 2% increase in ad prices, driven by the Asia-Pacific region and other regions. Growth was driven by advertiser demand and exchange rate fluctuations.
Other revenue from the Family of apps was $334 million, with a growth rate of 82%, mainly driven by the growth of WhatsApp Business messages.
RL revenue was driven by strong sales of Quest3 during the holiday season.
Outlook
(1) We have seen continued growth in Reels and overall video views, with daily average watch time for all video types increasing by over 25% in Q4, benefiting from algorithm improvements.
(2) One of the main focuses this year is to integrate videos within the FB ecosystem, unify the viewing experience and rankings, and create a smoother user recommendation system.
(3) Threads has shown strong momentum.
(4) At the end of the year, we launched our Meta Platforms AI assistant and other AI chat experiences in the United States. We tested over 20 Gen AI features in our apps. The focus for 2024 is to advance the release of Llama 3.(5) The driving force behind the growth of advertising is the improvement of monetization potential. On one hand, increasing the ad loading rate in highly engaging content for users, and on the other hand, improving ROI. The latter can be achieved through AI (Advantage+, Text Variations, Image Expansion) and deepening cooperation with enterprises to test results by integrating marketing data, thereby improving accuracy and conversion rates.
(6) Investing in new and engaging advertising experiences within the platform. For example, Click-to-Message is a focus area with strong growth. In addition to Messages, the advertising revenue under Shops has reached an annualized level of $2 billion this quarter.
(7) Capital allocation
Increase investment in the AI field, including computing power, by 2024. It is expected that investment will be more proactive in the coming years.
Reality Labs is a long-term investment plan, striving to develop the next generation of computing devices.
Our strong financial position and performance enable us to continue returning capital to investors. Previously, we did this through share repurchases, and now we will continue to maintain an active share repurchase plan while regularly paying dividends.
We will continue to monitor the regulatory environment, including the increasing regulatory resistance from the European Union and the United States, which may have a significant impact on our business. It is worth noting that the FTC is actively seeking to modify our previous consent decree, which would restrict some of our operations. We will make certain protests and appeals, but if we are not successful, it will have an adverse impact on our business.
(8) Revenue guidance for 1Q24 is $34.5-37 billion, assuming exchange rates remain unchanged.
(9) Operating expenses of $94-99 billion, with the main sources of growth coming from increased depreciation due to infrastructure costs, talent recruitment, and expanded investment in Reality Labs.
(10) Capital expenditures of $30-37 billion, with an increase of $2 billion at the upper limit. This is mainly driven by investments in servers, including AI hardware or non-AI hardware, data centers, etc.
IV. Analyst Q&A
Q1: What are the main factors driving the improvement in advertising effectiveness? Are there any areas where you continue to receive positive feedback on improvement? The acceleration in first-quarter guidance, what aspects or regions' performance give you confidence?
A: The main factor is the improvement in advertising conversion efficiency, which is the most common feedback from advertisers. The key is to enhance the advertising experience on the platform. For example, Shop ads were only opened to advertisers across the United States in the second quarter of last year, and now they have already achieved an annualized revenue of $2 billion. Secondly, we integrate advertisers' marketing data to facilitate adjustments to product experiences and generate better advertising conversion efficiency. Thirdly, we use AI to improve targeted ad predictions. AI tools such as Advantage+ Catalog, Advantage+ Creative, Advantage+ Audience, and so on.Regarding the confidence in guidance, it mainly reflects many trends we observed in the fourth quarter, namely strong and widespread advertising demand across various industries, especially in the online commerce and gaming sectors.
Q2: What operational company experiences did we learn from in the efficiency year?
A: The efficiency year is about running our business in a more streamlined and stable manner, which creates an environment and confidence for us to invest in projects for ten years or even longer.
Many people say that our streamlining is a short-term action, but it's not. I have truly started to change my mindset because we believe that a more streamlined company will operate better. Despite the challenges in personnel adjustments, maintaining business discipline actually improves the overall efficiency of the company.
There is a backlog of hiring demand for 2024, but the headcount added this year is relatively low compared to historical levels. We expect to maintain this trend for the foreseeable future.
Another reason for improving profitability is that we have enhanced our ability to respond to the changes in the next 5-10 years. There are currently many risk factors, whether it's geopolitical or regulatory risks, and the development path of technology is also full of uncertainties. We hope to have the capability to make large-scale investments when necessary and to flexibly set our long-term goals.
Q3: Is the Capex guidance range referring to the lower limit or the upper limit?
A: The increase in the upper limit already reflects our changing understanding of demand. This is a dynamic forecasting process with multiple factors causing uncertainty, but it is certain that AI investment will continue to increase in the coming years. Signs of increased AI investment can also be seen in 2024.
Q4: Previously, it was mentioned that it would take 10 years for the metaverse to be realized. Now that we have AI, has this timeline changed?
A: AI will make our products and services better. However, it is difficult to determine the exact development path. For example, when it comes to smart glasses, we used to believe that we had to build complete displays and holograms and provide a sense of presence for smart glasses to become mainstream products. But now, it seems that smart glasses with built-in AI assistants could become killer applications, and holograms and a sense of presence may come later.
Overall, I believe that these two aspects complement each other, and there may be some unexpected developments in the sequence of technological advancements.
Q5: How does Meta Platforms integrate with the shopping scene?
A: E-commerce is one of the strongest industries for our advertising business. Advantage+ has improved the ROI of DR advertising, and Ad+ has played a significant role in the growth of e-commerce advertising. We will continue to invest in this area.
In terms of specific application scenarios, for example, eligible Shopify businesses can now seamlessly join the stores on Facebook and Instagram. We are making it easier for advertisers to convert their existing ads into shopping ads, and we will continue to focus on deepening integration with partners and leveraging artificial intelligence to make shopping ads more efficient.Amazon Advertising's pilot program is one aspect of our collaboration with Amazon. The goal is to create a more seamless shopping experience on Facebook and Instagram through Prime purchases, making it easier for people to buy products directly from Amazon ads. This is still in the early testing phase.
Q6: Long-term growth opportunities for Messaging?
A: Regarding monetization of Messaging, we currently have two main avenues: click-to-chat ads and paid messages. We are investing heavily in both areas to facilitate customer engagement throughout the entire cycle. We have seen strong revenue growth in click-to-chat ads in the fourth quarter, with more and more advertisers adopting them.
These ads primarily cater to the marketing needs of advertisers at the lower end of the sales funnel (more direct brand marketing). We plan to optimize Click-to-WhatsApp during this year. Q4 data shows that Click-to-Message ads have remained strong, with more advertisers adopting them.
I would like to emphasize that our investment is focused on improving conversion and reporting accuracy. We plan to launch paid click optimization during Click-to-WhatsApp campaigns and provide customers with more comprehensive and solid reports to help advertisers understand how to engage users through click-to-messaging. In the future, if we can enhance automation and achieve interaction with customers through AI, there is room for further expansion.
As for paid message clicks, it is still in the early stages, but the data momentum is promising. The latest pricing model feedback has been positive since its launch in June. We will continue to lower the payment threshold and test the product experience.
Similar to click-to-chat, we are investing in making it easier to accomplish more tasks within the app. In October, we globally launched Flows, which allows businesses to provide a richer user experience on WhatsApp, such as selecting airplane seats or making appointments. It has shown good initial appeal, and we will focus on introducing more ways to help businesses easily create Flows and plan to introduce more features.
Q7: How do you view the competition and growth in the second half of this year?
A: We cannot share guidance beyond the first quarter. The full-year revenue will be influenced by various factors, including macro conditions, and the longer the time interval, the more difficult it is to predict. As you mentioned, as we enter 2024, we will also experience a cycle of increasingly strong demand. Therefore, we do not provide guidance beyond the first quarter. As the year progresses, we will have a better judgment on the next few quarters.
Q8: Reasons for dividend distribution?
A: We believe that introducing dividends can complement our existing stock repurchase program. The introduction of dividends does not change how much capital we will return or how we determine the total amount of capital to be returned.
We expect stock repurchases to continue to be the primary way we return capital to shareholders. However, introducing dividends provides a more balanced approach to our capital return plan and adds some flexibility in how we return capital in the future.Q9: Apple's decision to lift the download blockade on the App Store in Europe, allowing third-party channels, what role will Meta Platforms play in this?
A: Apple's move doesn't really affect us. I also believe that developers are unlikely to choose to enter the alternative app stores they provide. They have made the process very cumbersome, which is not in line with the purpose of the EU regulations, so it will be very difficult for anyone to seriously consider what they are doing.
Q10: Can you talk about the progress and monetization potential of business messaging on WhatsApp and Messenger?
A: GenAl products will not be a significant revenue driver in 2024. However, we certainly expect them to have the potential to become meaningful contributors over time.
Currently, the recent monetization opportunities lie in our advertising creative tools. These features, such as text variations and image extensions, have been widely rolled out and are now available globally. We have seen these tools being adopted even at this early stage and providing real value to advertisers. We will continue to invest in making these tools more useful and efficient, which will lead to more adoption by advertisers and create a virtuous feedback loop.
In the longer term, the GenAI capabilities we bring to business messaging do represent a compelling opportunity. We are currently conducting very small-scale AI chat tests with a few select businesses, but it takes time to make these AIs increasingly useful. We have received positive feedback from the businesses testing them, and we expect to make progress as we expand the testing scope this year.
On the consumer side, similarly, we believe GenAl can make it easier for people to create engaging content across all of our apps, including our messaging apps. Our AI assistants also provide additional utility. So in all these areas, we will follow the typical approach of testing and iterating until they reach a good level, and then we will invest in growth and ultimately explore commercial monetization.
Q: Which apps in the Family of Apps have seen increased engagement due to Meta Platforms AI? Will there be more commercial activities as AI agents or other AI tools are used more?
A: I think having a great assistant will be one of the true values created by this generation of artificial intelligence, and it also provides an opportunity for every creator to have an assistant or agent that people can interact with. The same goes for every business and agent, and it also allows people to create many quirky and interesting things. I believe that Meta Platforms AI will play a very important role in our products. It is currently available in WhatsApp, Messenger, and Instagram in some countries, but at this stage, we are not actively promoting it. As people use it, we are learning how they want to interact with it in fundamental ways.We started rolling out this feature around October or November last year, and it is currently in the adjustment phase.
This year, we expect to enhance this feature in our existing applications. We have conducted a series of experiments, launched products, received feedback, and finally carried out extensive integration. There were even jokes about us adding a story feature in the settings, although that never happened, but I found it interesting. We are currently in the learning and adjustment phase, and working on the next set of integrations, planning where these features will be available in the product. I am very excited about it, and it is also an important theme for us in 2024.
Risk Disclosure and Statement for this article: Dolphin Research Disclaimer and General Disclosure
