Meta AI Growth Opportunity

Meta Earnings Call

Notes & Figures

Meta’s stock jumped 9.83% in the first 6 hours after the beginning of their earnings call. Here are some statistics from the call I found interesting:

While ad impressions significantly increased, the decrease in price per ad suggests Meta's strategy focuses on increasing ad volume in platforms like Reels rather than solely on improved targeting (which would typically lead to higher prices).

Meta is turning CapEx into a revenue growth lever since the discovery that ad load elasticity in short-form video is far higher than expected, which means revenue can scale without increasing CPMs.

While clearly there is some investor concern in regards to cash flow allocation, I believe that user retention in the context of screentime-share with competing content delivery platforms will be important. Specifically, TikTok, which now serves American viewers content from an American-owned arm, should allow them to capture more market share and face less regulation.

Content consumption, not sure if I would call Instagram a social media platform ATP, is going no where; and I even expect its importance, relevance and scale to continue to grow exponentially over the next 10-15 years.

Scalable Recommendation Model... a Novel Use of Transformers

The Generative Ads Model (GEM) stands as the most compelling justification for this massive spending.

Reality Labs, Ray Band Glasses & Genie 3

Despite significant investment, Reality Labs reported a loss of $6.02 billion in Q4 2025, with revenue down 12% YoY. Historically, Wall Street has punished Meta for Reality Labs investment; but Mark Zuckerberg’s willingness to keep his foot on the gas with Reality Labs suggests he expects a coming shift in technological capability that would enable Reality labs to capitalize on many years of investment.

I think that we're at a moment similar to when smartphones arrived... It's hard to imagine a world in several years where most glasses that people wear aren't AI glasses

Apple Vision Pro

Apple Vision Pro & all AR/VR Meta ventures mean that, despite all of the doubt and haters, the largest hardware & media incumbent respectively believe in this technology. Meta’s 2014 acquisition of Oculus may have been slightly early in terms of product-market fit, but gave them access to the bleeding edge of the talent and technology at the time.

Essentially, I argue their acquisition is best justified in the context of today; where large incumbents are buying up talent, users and market expertise to carve out well defended verticals in fields likely to become more relevant in a new era redefined by AGI & other technological advancements.

By acquiring companies in related industries to their fundamental product, these companies gain the experience and knowledge of experts in specific niches. This enables them to built out verticals that enhance their fundamental product offering without taking team bandwidth away from current focuses.

Some other things to consider:

The phone/ computer wont last forever… or at least I hope not

Carrying a phone/computer around leads to a user-interface disconnect in that phones, at best, occupy approximately 5% of a user's field of view (FOV), while computers might cover around 20% and users can’t look at their device 24 hours a day.

As ads begin to become more embedded in digital interactions and harder to distinguish from normal content, as a firm, I now have to compete for attention with the rest of the users FOV, in and outside of the device.

I urge readers to go demo Genie 3 if possible, or just scroll through X for 30 seconds. It is very clear to me that this technology, in conjunction with VR glasses is the future of our media consumption.

Genie 3 marks a significantly large step for in world model generation

Google officially released Genie 3, their world generation model, to US AI Ultra subscribers after announcing it ~six months ago. The newest videos I’ve seen are much more impressive than what was shown in the official release. World models offer many different important use cases including:

This technology threatens to disrupt many current domains and I am very excited to watch scaling efforts continue. In an end of year podcast, Demis Hassabis talked about fusing or merging their video model Veo with Genie. Although I am unsure as how this would work, is Veo utilizing Genie for physics adherence or Genie utilizing Veo for ideation; the prospect of this excites me both from a nerdy architecture perspective but more importantly, for the future of this domain.

Fei Fei Li, Yann Lecun among others have talked about the importance of world models. Fei Fei Li created world labs, which offers a similar product offering to Genie but instead uses Gaussian splatting to generate worlds. Currently their technology is far surpassed by current Genie 3 and likely the checkpoint of when Genie 3 was announced as well.

In this Dwarkesh Patel episode, Richard Sutton (the Godfather of AI) seems to believe fusing LLMs to world models offers an interesting path to AGI. Although I can’t pretend to understand how this would work whatsoever, I too am excited about the future of world models.

Future of Advertising & Content Consumption

It is my belief that generative AI will soon usher in an age of advertising like that of which we have ever seen. I see the potential for embedded ads into content, for example a can in the background is changed from Coke→ Pepsi → Sprite depending on user preferences, ad supply and companies intended customers. I think content customization and personalization with GenAI will begin to make huge impacts on the field.

To learn more about this concept, checkout a paper I wrote at: https://www.maxwellmoroz.com/adpipelinepaper.pdf or to see a (very basic) demonstration of how this technology might work go to: https://adembeddings.vercel.app/feed.

Here is a screenshot from the site, the same image may be served to two different users, but one may show Monday.com on the screen, while the other shows a Clickup ad.

Other exciting MetaAI Technology

Tidbits

A conventional narrative you might come across is that AI is too far along for a new, research-focused startup to outcompete and outexecute the incumbents of AI. This is exactly the sentiment I listened to often when OpenAI started ("how could the few of you possibly compete with Google?") and 1) it was very wrong, and then 2) it was very wrong again with a whole another round of startups who are now challenging OpenAI in turn, and imo it still continues to be wrong today.

- Andrej Karpathy on https://flappingairplanes.com/

Thank You so Much for Reading this Far

I am so incredibly grateful for you taking the time to hear my thoughts and perspective, PLEASE reach out to chat about any of this. We are in an incredibly exciting time and I am always excited to hear more about anything.