(Speedwell Members click here for a longer version of this post or here for a PDF version of this post here)
4Q25 Update.
Meta reported 4Q25 earnings and the stock popped +7% after hours.
Their core business continues to perform very strongly with top line revenues up +24% y/y. While that is a slight 200bps decel from last quarter, it still is impressive.
Daily active people grew +7% y/y to 3.58 billion.
Ad impressions grew +18% y/y, which was an 800bps jump from last quarter’s +10% y/y.
Ad pricing increased by +6% y/y, which was 400bps less than last quarter. This is a positive as the better Meta is with ad targeting, the more ad pricing decreases initially, which improves ROAS for advertisers.
While the business is performing well, their spending has exploded. Total costs and expenses are +40% y/y and operating income was up only +6% y/y.
In 2026 they anticipate another huge jump up in expenses, with guidance calling for $162-169bn in costs, up about +41% y/y at the midpoint. Part of this is because of an increase in depreciation (as readers will recall we anticipated this over the past several write-ups), and also growing infrastructure costs (again).
They also note that the second-largest contributor to total expense growth will be employee compensation, which includes AI hires. (Next year will be the first full year of all of those very high AI researchers pay packages that grabbed headlines).
Unsurprisingly, investors should not expect much earnings growth next year. Meta’s press release noted 2026 operating income would be “above” 2025, suggesting barely any growth.
The discrepancy between cash flow from operations and free cash flow is as large as it’s ever been. On the one hand they have generated an impressive $95bn in operating cash flow in 2025 (after subtracting SBC and despite Reality Lab losses). However, after capex and financing lease payments, free cash flow drops to just $23bn.
Capex is expected to increase even more next year to $115-135bn, meaning free cash flow will likely all but disappear.
Business Commentary.
Zooming out, Meta is killing it. They seem to be one of the clearest beneficiaries of AI and every stat on the call mentions how AI is improving time spent or ad conversion. Instagram Reels watch time was up +30% y/y in the United States. Facebook optimizations resulted in a +7% lift in organic feed and video posts, which resulted in the largest quarterly revenue impact from Facebook product launches in two years. A new ad attribution model drove a +24% increase in incremental conversion vs their standard attribution model. Their “GEM” model drove a +3.5% lift in ad clicks on Facebook and a +1% gain on Instagram. And adjusting the timing of when an ad is shown on Facebook drove a 4x larger revenue impact than just increasing the ad load, suggesting ad optimization can continue on a new vector besides matching the right ad to the right person.
Even outside of content recommendations and ad targeting, AI is benefiting the actual ad copy. Their video generation tools hit $10bn in revenues in 4Q, with q/q growth outpacing the increase in overall ad revenue by nearly 3x, suggesting this will be a new driver to increase advertisers’ willingness to spend.
There also is AI agents, which they think will help them build new products and work more efficiently internally. They noted that they saw a 30% increase in output per engineer with agentic coding.
Zuckerberg also talked about how AI can change the app from an algorithm to a more personalized experience. He was sparse on details (and historically there haven’t been too many changes to the core app that they have invented that users appreciated. The original Feed idea was a Facebook invention, but Stories and Reels were copied).
Zuckerberg is really trying to tie together their family of apps, AI, and their Reality Labs efforts. He noted that: “Glasses are the ultimate incarnation of this vision” with automatic data collection feeding an AI that can create a custom UI for the user. Zuckerberg seems so impressed with his own vision that he commented: “It’s hard to imagine a world in several years where most glasses that people wear aren’t AI glasses.”
While the glasses are more popular than Apple’s VisionPro and they are best positioned for the shift to AR/VR, the existing products have yet to find a compelling use case for consumers with lackluster retention. Of course that could change… and Meta has bet close to hundred billion believing it will.
For the first time this quarter they noted that Reality Lab losses will start to go down after next year (starting in 2027). This will no doubt be more than welcomed by investors, but the $20bn in RL losses are quickly overshadowed by their expenses expected to increase beyond $162bn (at the low end) next year.
The real call option is not Reality Labs, but AI. They are spending as much if not more than any other player to research and build infrastructure for AI. While they have been largely lacking from the AI race leaderboards, Meta spent an incredible amount of money to attract top talent last year to change that. Zuckerberg expects that in the next few months they will start to ship new models. He doesn’t expect the first model to be revolutionary but believes the system (and money) they have in place will allow them to reiterate faster than competitors to eventually push them to the frontier.
It doesn’t seem like many investors (or AI researchers) necessarily believe that, but that would be the real blue sky scenario. If such an outcome happened, they could spin up a new consumer app or increase engagement on their existing family of apps with deep AI integration. Although, out of all the big tech players, it is hard to see people trusting Meta with all of their personal information, and even if they have a better model, competitors have already built a habit with users regularly using ChatGPT and Gemini. That only means that the bar has risen even more for them to displace existing usage.
However, they don’t need a knock out consumer product to be successful here. Even creating better models that are lighter (require less compute and thus money) could be a big win for them. Right now they can’t deploy their biggest models for ad targeting and have to create lighter ones. These are cheaper to use and have a better ROI, but are not as effective.
While AI is no doubt benefiting Meta and they are getting an ROI with revenues reaccelerating to 20%+ from -1% in 2022 and +16% in 2023, it is worth remembering that expenses increased +40% in this quarter (+25% for 2025) and it is unclear how much of this was needed to support this growth. While we suspect a majority of these expenses are “growth” and not “maintenance”, which implies they are not ongoing, it is possible that pinging all of these AI models costs more than their legacy machine learning models. They may be driving revenue growth, but the cost to get that revenue growth is going up too. FOA margins were 60% in 4Q24 versus 52% this quarter—and that is before a lot of the elevated capex runs through as depreciation.
That is not to say that it is a bad ROI or a poor investment, but just something to be aware of. Said another way, they had to increase investment if they wanted to continue to grow. We just don’t know yet how much of this investment can be curtailed with them keeping much—or all—of the benefit.
To understand what is priced in and the growth an investor needs to assume in order to make a return, we turn to our Reverse DCF.
Valuation.
The rest of this update is only accessible to Speedwell Members.
If you are a Speedwell Member, click here to read the rest of this post.
If you are not currently a Speedwell Member, but wish to join, click here.
If you’d like to purchase the individual Meta Research Report and get access to the full 166+ page report and the full updated reverse DCF section, click here.
For further reading, check out our Meta Extensive Research Report here.
The Synopsis Podcast.
Follow our Podcast below. We have four episode formats: “company” episodes that breakdown in-depth each business we write a report on, “dialogue” episodes that cover various business and investing topics, “article” episodes where we read our weekly memos, and “interviews”.
Speedwell Research Reports.
Become a Speedwell Research Member to receive all of our in-depth research reports, shorter exploratory reports, updates, and Plus members also receive Excels.
(Many members have gotten their memberships expensed. If you need us to talk with your compliance department to become an approved vendor, please reach out at info@speedwellresearch.com).
*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.
Leave a Reply