(Speedwell Members can find a PDF version of this post here)

2Q25 Earnings Update.

Meta reported a very strong 2Q25 and the stock popped +10% after hours.

Revenue reaccelerated +22% y/y, up 600bps from last quarter’s +16%.

Ad impressions grew +11% y/y, mainly driven by Asia Pacific. Average price per ad increased +9% driven by increased advertising demand and improved ad performance. CFO Susan Li noted on the call that “pricing growth slowed modestly from 1Q due to the accelerated impression growth in Q2”, which is exactly what you would want to see.

Operating income margins expanded to 500bps y/y to 43%, which includes the reality labs loss of ~$4.5bn this quarter.

FOA (Family of Apps) generated $25bn in operating income for the quarter, +29% y/y. FOA Operating margins expanded ~350bps y/y to 53%.

Diluted EPS for the quarter was $7.14, +38% y/y.

Zuckerberg noted on the call that “there are 5 basic opportunities that we are pursuing: improved advertising, more engaging experiences, business messaging, Meta AI and AI devices.” Their strong execution is largely due to AI, which was credited with unlocking greater efficiency across their ad system. Their AI-powered ad recommendation model drove 5% more ad conversions on Instagram and 3% on Facebook.

He also noted progress with generative AI that improves ads on advertisers behalf. This is going to be especially helpful for smaller advertisers who don’t have large budgets to spend on professional creative.

In terms of more engaging experiences, improved content recommendations have driven a 5% increase in time spent on Facebook and 6% on Instagram in 2Q alone. New AI editing tools can also improve content. On the call Susan Li noted that video growth was up 20% y/y. Within that she said that 2/3rds of recommendations are now original content (as opposed to reposts from TikTok, which was a common criticism of Reels). She said they expect further improvements this year.

The other opportunities he mentions are less impactful to the financial results today. Zuckerberg expects every business to one day have a Business AI the same way they have an email address and this will enable more business messaging. As they have the most businesses of any platform already onboarded, and together with Messenger and WhatsApp, they are well positioned to benefit from this.

Meta AI has a billion users who access it through their apps. He intends to make Meta AI the leading personal AI assistant. While the focus on an AI assistant may seem like a non sequitur for a social media company, there is some rationale. First it could improve the app experience and prevent people from leaving when they have a question.  Idealistically, a lot of the activity people use ChatGPT or Google for, they could use Meta AI for. This would increase time spent on their apps.

Meta could have some unique advantages here: if you wanted to search for restaurant or beauty recommendations, surfacing a bunch of Reels could be a better experience than text or listings. They could also use the data they have on you, your friends, and your interest to better surface results.  

While Zuckerberg probably hopes that they can have another consumer app (or at least product) hit, this is going to be hard given OpenAI already has a very popular consumer product and Google has many consumer touchpoints. In the current AI landscape, and given consumer trepidation of trusting Meta with too much data, they are facing an uphill battle.

Another possibility is that Zuckerberg is getting paranoid by these new AI services that are getting direct customer touch points. With enough advancement in AI-content generation, it may be possible for someone to just request the sort of content they wish to see and thus turning every AI app into de facto entertainment destination. Entertainment is one of the biggest reasons people end up on Instagram or Facebook. While this also seems a bit farfetched, there is no doubt that there will be a time spent war for consumer’s limited attention.

Sam Altman likes to point out that people don’t tend to regret the time they spend on ChatGPT as they do with social media. While a social media habit has proved to be sticky for billions of people, it is possible that consumer preferences change overtime—which could be a huge headwind for Meta. Perhaps Meta having their own AI assistant and app could be a hedge against that.

More clearly though, the AI assistant plays into their efforts to build a leading AR/VR platform, which relies on AI for the consumer to interface with it. This quarter they noted strong momentum with Meta’s Ray-Ban glasses. Zuckerberg sees the glasses as a key way that people will integrate “super intelligence” into their everyday lives.

Business Commentary.

Zooming out, everything is going well for Meta. Usage is up (Daily app users +6% y/y), time spent is growing, and their ability to target ads is improving. Even more incredible is that there doesn’t seem to be an immediate limit to how much AI can help improve recommendations of content and ads, in addition to it helping improve both with Generative AI and AI-enabled editors. They are very well positioned to benefit from AI improvements, even if their Llama model never catches up to OpenAI or Google.  

While it is still unclear how much of a reality their Reality Labs efforts are, they will not lose ~$18bn a year indefinitely. An investor can assume that there will either be real value created or they will shutter the division. It doesn’t seem likely that they will continue to spend at this level 10 years from now if they still haven’t gained broad adoption of their devices. While 10 years may be a long time, the fact that it is not forever is an important consideration when thinking of the DCF.

The only real question that lingers over Meta is their normalized free cash. After adjusting for SBC and backing out their massive $16.5bn quarterly capex spend, it leaves Meta with just $4.2bn in free cash flow. This is down ~35% y/y from $6.6bn. This yields a very low 23% net income to free cash flow conversion figure. (For each dollar of net income they report, only 23 cents are actual cash flow).

Now of course this is because of their elevated capex investment into servers, data centers, and other infrastructure. But their capex to depreciation ratio increased to 3.8x this quarter from 2.2x last year. We have constantly been drawing attention to this because 1) mathematically it means that depreciation is going to continue to rise and 2) it is likely earnings are overstated.

Remember that depreciation is nothing more than a cost allocation method. The idea is that a business matches the cost of using its assets in a given period to correspond to the revenues that are generated in the same period.

Investors often make the distinction between “maintenance” and “growth” capex because a portion of capex in a period can be just to maintain existing operations, and another portion is directed towards growth—which doesn’t show up in the current financials. It is typically hard to estimate this, but it is still important because when you strip out the portion of “growth capex” you can better understand and value a business’s normalized cash flow.

Now with Meta we know that their capex is running wildly ahead of their depreciation. Because of the way accounting works though, the depreciation captured is backwards looking. The question now is does Meta need to spend ~$70bn a year in capex just to maintain their existing business or is it closer to their current depreciation run rate of $17bn?

While no doubt their capex is helping them grow as 2022 revenues were $116bn and they are on track to do close to ~$190bn for the year. Most recent earnings calls are flooded with commentary about AI-driven improvements in recommendations and ad targeting (which was supported by their capex spend ). Over three years that is ~$75bn in revenue growth compared to ~$135bn in cumulative capex spend.

For the 10 year period prior to them embarking on this elevated capex spend (which started in 2023), capex as a % of revenue averaged 17%. If we assume that they could have run at this level and maintained that $116bn in revenue they had in 2022, then we can back into growth capex.

Assuming maintenance capex was 17% of revenue, that implies normalized depreciation of ~$20bn. This leaves $115bn of their past 3 year capex spend as growth capex. If we take their incremental revenue growth of $75bn and apply a 50% margin, that is an incremental after-tax ROIC of 25%. While this is a great ROIC, especially considering the scale of capital, it is likely that it is really higher. This is because it can increase by making better models with AI advancements that will not necessarily require more servers.

However, this still leaves the question of what their new “maintenance capex as a % of revenue” figure would be since they transitioned the business to a more AI-dependent and compute intensive model.  While these data center build outs will include a lot of one-time expenses that will be cheaper to maintain than build from scratch, their normalized D&A is still likely much higher. If their new normalized deprecation was twice what it is today, that would be a $5 hit to EPS. 2024 EPS was about ~$24, so this is far from a trivial point.

The tricky part though is that they can possibly continue to grow even more just off of their existing assets… perhaps enough to fully offset that.

High level, Meta tends to generate about 46-48% of their revenues in 1H. If we apply this same seasonality to their EPS, that is a 2025e EPS of $28.25-$30. At Meta’s current stock price of $775, that implies a ~26x multiple. If we back that $5 estimated depreciation adjustment, that multiple jumps to 32x. However, we also have to remember that there is about $5.60 of Reality Labs losses per share embedded in those earnings. If you back those out, then it more than offsets the depreciation adjustment.

While every investor will need to make their own judgement on normalized cash flow and what Meta’s long-term growth prospectus look like, a 26x multiple can easily be rationalized by a couple more years of high teens to 20%+ growth. It may not exactly be the most conservative assumption, but just 1.5 year of 20% growth brings a 26x multiple to 20x.

Meta Call Notes.

Margin Pressure

• “While we’re still very early in planning for next year, there are a few factors we expect will provide meaningful upward pressure on our 2026 total expense growth rate. The largest single driver of growth will be infrastructure costs driven by a sharp acceleration in depreciation expense growth and higher operating costs as we continue to scale.”

Unparallel Compute

• “The people who are joining us are going to have access to unparalleled compute as we build out several multi-gigawatt clusters.”
• Prometheus coming online next year and they think it will be the world’s first gigawatt-plus cluster
• Hyperion will take a few years, but will scale to a 5 gigawatt cluster. They also have other “Titan” clusters in development.

WhatsApp

• Family of Apps other revenue was $583 million, up 50% driven by WhatsApp paid messaging revenue growth as well as Meta Verified subscriptions
• WhatsApp is rolling out ads and status and channels, along with channel subscriptions and the updates tab to help businesses reach the more than 1.5 billion daily actives who visit that part of the app.
• “WhatsApp ads and status to earn a lower average price than Facebook or Instagram ads for the foreseeable future, due in part towards WhatsApp skewed toward lower monetizing markets, and more limited information that can be used for targeting. Given this, we do not expect ads and status to be a meaningful contributor to total impressions or revenue growth for the next few years”

Threads

• “Advertisers globally can now run video and image ads to Threads users in most countries, including the United States. While ad supply remains low and Threads is not expected to be a meaningful contributor to overall impression growth in the near term, we are optimistic about the longer-term opportunity with Threads as the community and engagement grow and monetization scales.”

Reality Labs

• Q2 revenue was $370 million, up 5% y/y due to increased sales of AI glasses, partially offset by lower Quest sales.

Ad Efficiency

• “First is our Ad systems where we’re innovating in both the ads retrieval and ranking stages to serve more relevant ads to people. A lot of this work involves us continuing to advance the modeling innovations we’ve introduced previously while expanding their adoption across our platform.”
  • The Andromeda model architecture powers the ads retrieval stage of our ad system, where we select the few thousand most relevant ads from tens of millions of potential candidates.
    • In Q2, we made enhancements to Andromeda that enabled it to select more relevant and more personalized ads candidates while also expanding coverage to Facebook Reels. These improvements have driven nearly 4% higher conversions on Facebook Mobile Feed and Reels.           
• Our new Generative Ads Recommendation System, or GEM, powers the ranking stage of our ad system, which is the part of the process after ads retrieval where we determine which ads to show someone from candidates suggested by our retrieval engine
  • “ In Q2, we improved the performance of GEM by further scaling our training capacity and adding organic and ads engagement data on Instagram. We also incorporated new advanced sequence modeling techniques that helped us double the length of event sequences we use, enabling our systems to consider a longer history of the content or ads that a person has engaged with in order to provide better ad selections.”
• “Seeing strong momentum with our Advantage+ suite of AI-powered solutions.”
  • “We’ve seen lifts in advertiser adoption of sales and app campaigns since we’ve expanded availability, and are working to complete the rollout for leads campaigns in the coming months.”
  • “Nearly 2 million advertisers are now using our video generation features, image animation and video expansion, and we’re seeing strong results with our text generation tools as we continue to add new features.”

AI System Improvements

• “Over the last few months we have begun to see glimpses of our AI systems improving themselves. The improvement is slow for now, but undeniable. Developing superintelligence is now in sight.”
• “Meta’s vision is to bring personal super intelligence to everyone, so that people can direct it towards what they value in their own lives.”

Expenses

• While we’re still very early in planning for next year, there are a few factors we expect will provide meaningful upward pressure on our 2026 total expense growth rate. The largest single driver of growth will be infrastructure costs driven by a sharp acceleration in depreciation expense growth and higher operating costs as we continue to scale.

Capex

• We currently expect 2025 capital expenditures, including principal payments on finance leases, to be in the range of $66-72bn.”
• “The big driver of our increased CapEx in ’26 will be scaling GenAI capacity as we build out training capacity that’s going to drive higher spend across servers, networking, data centers next year. We also expect that we’re going to continue investing significantly in core AI in 2026.”
• Open to codeveloping data centers with partners to help share 2026 capex outlay

Guidance

• We expect 3Q25 total revenue to be in the range of $47.5 to $50.5bn. Guidance assumes foreign currency 1% tailwind
• We expect full year 2025 total expenses to be in the range of $114 to $118bn, narrowed from our prior outlook of $113 to $118bn and reflecting a growth rate of 20 to 24% year-over-year.

A Meta Data Center (Generated by AI)

---

*At the time of this writing, one or more contributors to this report has a position in Meta. Furthermore, accounts one or more contributors advise on may also have a position in Meta. This may change without notice.