Meta is all in on AI. To get a sense of how aggressive the company is pursuing the paradigm shift, we tracked the number of times “AI” or “machine learning” were mentioned in the company’s prepared remarks and, as expected, found the pace of those mentions have been quickening.

• 2019: 0
• 2020: 1
• 2021: 12
• 2022: 64
• 2023 (first three earnings calls): 97

That work yielded new AI product announcements at the company’s September 2023 Meta Connect developer conference, most notably with Meta AI, EMU, and “Celebrity” AI Assistants. All of these products will increasingly drive engagement in 2024 but will have little if any effect on revenue this year, because they won’t start aggressively monetizing until 2025 or beyond. 

• Meta AI (available today through Facebook messenger, Instagram chats, and WhatsApp): When you compose a message in Facebook Messenger, Instagram DM, and, eventually, WhatsApp, an option now appears at the top of the stack, “Meta AI.” This is basically a slimmed-down version of ChatGPT. It’s a question-and-answer chatbot based on Llama 2 (I’ll cover Llama later on), with real-time information capabilities driven by a partnership with Microsoft and Bing search.
  • My take: While the product feels out of place sitting at the top of the message feed, its placement inside Meta’s messaging apps will get the product in front of more than a billon people per month.
• EMU, or Expressive Media Universe (still in development): EMU is a new AI image creator that is similar to DALL-E and Midjourney. Users can prompt EMU with any imagery, like “astronaut underwater,” and receive an AI image back. The bigger feature is EMU image editing within Instagram. Users will be able to use EMU to edit their pictures by asking for edits via prompts. Zuckerberg gave an example in the keynote of a user asking EMU to edit a picture of their dog into an “origami” creation, making the dog appear more animated and as if made out of folded paper.
  • My take: While the product is not available today, I believe this feature will be a hit and change the face of Instagram. Today Meta has a teaser image generation product that is available, “Imagine with Meta AI,” which is a better version of DALL·E 3, and not as advanced as Midjourney.
• “Celebrity” AI Assistants (available today through Facebook messenger, Instagram chats, and WhatsApp): AI chatbots loosely based on individual celebrities designed for specific areas of interest. For example, they showed a character “Bru,” the go-to sports guru for all sports-related questions or conversations that’s portrayed by Tom Brady.  By targeting a wide audience with interests ranging from cooking to medieval roleplaying in an innovative entertainment format, user time spent will likely increase.
  • My take: Meta is experimenting with AI-powered content formats. While I expect the concept of the “Celebrity” AI assistant to fizz out, I believe the concept of generative AI avatars, based on real and fictional characters, will be a hit. The reason is Instagram is about impressive media and lends itself to experimentation and adoption of new content formats.
• There was a fourth AI product announcement in September, Ray-Ban Stories Smart Glasses. These are basically an AI version of Snapchat’s “Spectacles” from 2017.
  • My take: This is basically a gimmick and likely won’t go anywhere.

Meta’s foundation model is called Llama, and Llama 2 is the latest version. It is arguably the best and most widely used open-source model on the market today. Open-source means the model is available to the public for use at virtually no cost. The company does charge hyperscalers, including Microsoft and Amazon, which offer access to Llama 2 as part of their own generative AI enterprise services at an undisclosed rate. This approach is different from the other leading models from OpenAI (GPT), Anthropic (Claude), Google (Gemini), and X.ai (Grok) that are closed-source/proprietary.

Since Meta does not charge a traditional software license, subscription or tokens, the central question is how and when will they make money from Llama 2?

Meta’s intention initially is not to focus on making money but instead to get the model into as many hands as possible to advance AI. On Meta’s June earnings, CEO Mark Zuckerberg said Llama would not generate “a large amount of revenue in the near term, but over the long term, hopefully that can be something.” I believe Zuckerberg is taking the one-billion user approach to building a business around Llama: Keep growing the base and figure out how to make money down the road. My prediction is in 2025 Meta will adopt a similar model as OpenAI and charge for tokens or use a subscription-based approach for access to the Llama model.

Zuckerberg thinks about the business in terms of five-year increments. I see them in these chapters:

• 2007 to 2011:  Growing Facebook outside of college and universities
• 2012-2016:  Growth of mobile
• 2017-2022: VR, AR, and the metaverse
• 2023: The year of efficiency
• 2024-2030: AI

It’s hard not to get caught up in the excitement of AI. With that excitement comes increased expectations as to how the paradigm shift will impact revenue. This year many tech investors will be wanting to see more contribution from AI, and with that there will likely be letdowns. It happened to Adobe late in 2023 when they first talked about 2024 revenue contribution from AI being below investor expectations.  I see that risk with shares of Meta in 2024 and encourage investors to stay the course.

Zuckerberg is looking to build a business with a billion-plus monthly users. He is not in this to increase revenue growth by a few percent in a given year to beat expectations. He is playing the long game. Just look at the nearly $40B he has invested in Reality Labs or the two years he has spent slowly ramping the monetization on Reels. I believe he will take the same approach with AI, announcing lots of new products that don’t generate meaningful revenue for years, and that’s okay.

The reason why it’s okay? I believe in three to five years we’re going to see an AI bubble that will rival the dot.com bubble. If that plays out to plan, the Nasdaq will 4x in the next five years. If Zuckerberg sticks to his time-tested approach to testing, releasing, ramping, and monetizing products, he will find another gem in AI that will drive a billion new users to the company and expand its market cap along the way.

Disclaimer