Meta’s business is going strong. Revenue growth in March was up 27% y/y, and the midpoint of June’s guidance calls for 18% growth.

Even more important than revenue growth in a given quarter, the best metric to track the core health of the business is related to daily users. The company reports Family Daily Active People (DAP or the number of people who use a Meta app on a daily basis), which grew 7% y/y in March to 3.2B people. That was basically in line with the 8% growth in Dec-23. We view any growth of DAP as progress given to grow that number in the first place is difficult because Meta is already used by 40% of the world’s population daily. We believe most of the DAU growth is coming from the Rest of the World (13% of revenue) and Europe (23% of revenue). DAU growth in the US is largely coming from WhatsApp, per Zuckerberg’s comments.

On the AI product front, the company continues to move forward quickly. Recently, they released Llama 3, integrated AI into user interfaces, released AI tools for advertisers, and announced plans for Agentic AI. Agentic, or agent-based, AI is the concept that a user can prompt a model once and the model will complete a series of tasks on its own. Some companies refer to this concept as personalized AI. While it’s a few years out, agent-based AI will represent a step function in utility compared to the current generative AI use case.

As mentioned, what concerned investors was Zuckerberg’s commentary that the company is entering into a multi-year investment cycle to build AI compute and pay for the corresponding energy. As part of entering this AI build-out phase, the company’s guidance for capex increased at the midpoint to 12% for 2024, which likely means it will carry forward into 2025 and possibly 2026 as well.

To better understand investors’ negative reaction to the outlook, it’s worth noting that over the past 15 years, Amazon has led the way in conditioning investors to stay on the sidelines during the first half of an investment cycle when margins decline, and return for the second half just before margins begin to rise.  As for Meta, operating margins are at a 3-year high and will likely decrease until some of the revenue growth effects of AI begin to materialize, likely late in 2025.

We believe when it comes to these investment periods, the investor should consider the context.  In the case of Meta, the company is in the early days of what will be a decade-long journey to infuse AI across their products, as well as launch new products (agentic AI and pay for Llama, just to name two). Collectively, the opportunity around AI is an order of magnitude bigger than the benefit that mobile has been over the past decade. During that period, annual revenue increased from about $8B (2013) to $134B (2023), or 17x. While the law of large numbers makes it next to impossible for overall revenue growth to match the past decade, on an absolute basis we believe the company will add more than $150B in revenue over the next decade.

What was clear on the call is Zuckerberg’s belief that he expects AI transformation to strengthen everything.

He outlined 4 key areas where the company can benefit from AI over the next decade:

1. End Products: Users of apps (Facebook, Instagram, Threads, WhatsApp) will interface with more AI features and recommendations, including Meta AI (the AI chatbot) which was recently rolled out as Instagram’s search function, and Facebook. Our take: this will have a slow adoption with users as they are used to searching for people, places, and things, unlike Q&A queries one would ask ChatGPT or Gemini.
2. Advertiser Tools: AI tools of advertisers allow for campaigns to be built quickly and cheaply. AI will help better understand when a user actually has an intent to buy and flag that event as an advertising opportunity. Advertisers, who account for 98% of Meta’s revenue, will have tools to improve ad efficiency, opportunity, and costs. Big picture: while the market reacted negatively to the March quarter, this is the right play for Meta.
3. Llama: Meta has big ambitions for Llama, and rightfully so given outside of Google they have the best AI training data. Currently, Llama is a free product and while the company did not give any indication that they will be charging, we believe it’s a function of time before the pay meter is turned on. Monetization could take two forms: 1) a traditional model of selling tokens, or 2) an AWS-like infrastructure arrangement where Meta builds compute, hosts Llama on the platform, and charges for access.
4. Agentic AI (Agent-based AI or Personalized AI): This represents the second phase of consumer AI, a step beyond the current generative use case. Current models operate like a conversation where you prompt the model and it gives a response. Agentic AI will take a single prompt and run several queries behind the scenes to complete the tasks. Zuckerberg gave examples of doing a research project or buying something. Additional examples could be scheduling a meeting, contacting a utility about a support bill, planning a trip, or doing just about anything that you do online today.

As mentioned, we believe the Agentic AI opportunity is significantly greater than the Generative AI opportunity. While Meta has the data and distribution (+3 billion users) to become a leader in agent-based AI, the challenge is convincing consumers that their data will be safe. We expect to hear more about Meta’s personalized AI opportunity in the coming quarters and could see a product in late 2025.

Disclaimer