Sections d, e, f & g are for paid readers. They get into 2026 guidance, valuation, CapEx, the balance sheet, conference call notes and my take.

Upgrade

In case you missed it:

The Big Beautiful Bill (BBB) implementation led to Meta writing down the value of some deferred tax assets. This led to a $16B non-cash charge, which heavily impacted EPS and led to a massive and noisy “miss.” Excluding this charge, EPS would have been $7.25 vs. $6.71 expected for 17% Y/Y growth. EPS is irrelevant for this quarter. The EPS growth chart below excludes this item.

2026 expense guidance was raised. The OpEx language was largely the same. It changed language from 2026 rates “above” 2025 rates to 2026 rates “significantly faster” than 2025 rates. Subtle, but I think it can still be taken as rising expense expectations. This also means faster than 22% Y/Y OpEx growth, which will lead to contracting margins in 2026. That is already expected.

Employee compensation from their hiring spree and more depreciation from data center buildouts are the main culprits. It also raised 2026 CapEx dollar growth guidance from “stable” to “notably larger.” This means CapEx will be at least $103B vs. $97B expected. That’s despite offloading some CapEx dollars as part of the Blue Owl financing deal. We’ll get into why it’s spending so aggressively in a moment. It also sounds like they’ll use 3rd-party partners for some of the financing (like the Blue Owl deal) and capacity growth. They may take out more debt too:

“The strong financial position and cash generation of our business enable us to make these investments while also accessing additional pools of cost-efficient capital.” – CFO Susan Li

Analyst profit estimates are probably going to fall for 2025 due to the tax charge. Given the expense commentary, I think estimates will modestly fall for 2026 as well. Meta trades for 26x EPS. EPS is set to grow by just 7% next year and then compound at a 13% clip over the next two years. It also trades for 60x forward FCF as it leans into data center CapEx. As of right now, FCF is set to shrink in 2025 and 2026, before compounding at a multi-year clip over 50% starting in 2027. Investment cycle.

Spending on AI & Superintelligence:

So why is Meta spending so aggressively? Is this 2022 all over again? As a reminder, Zuck spooked investors 3 years ago when he seemed determined to spend Meta's entire balance sheet on the metaverse. The company rightfully changed course, re-allocated investments and delivered a fantastic recovery. Some think the lofty 2026 expense guidance, which is powered by its superintelligence team and will pressure near-term margins, is Meta making the same mistake. They think they’re repeating the error of allocating too much capital to a speculative area. I could not disagree more strongly for two connected reasons. 

First, it has so much flexibility to allocate potential excess compute to different parts of its core business. They have not come close to using all of the compute they could within FOA, so overbuilding for ambitious goals like Superintelligence would simply mean that gets used elsewhere. Secondly and relatedly, the other use cases Meta has for this asset are already powering a large portion of its profitable growth engine. As we’ll discuss throughout this piece, Meta has a massive list of ways to use better foundational models and more compute to improve the two things that matter most for its business: engagement and monetization.

So... they have a lot of options for using future capacity... and many of those options are proven to come with high ROI.

They’ll continue to allocate capital in a way that maximizes use case flexibility so that they don’t get stuck with a large chunk of unneeded, decaying data center assets. Whether that’s building data centers in phases or using partners for some capacity needs, they want to remain optimally malleable. And while that’s the right decision, I think the risk of massive waste here is low for reasons already mentioned. Their optionality greatly diminishes risk of these investments coming with poor returns – regardless of how quickly superintelligence comes. And if superintelligence is reached sooner than expected? They want their expensive team of AI stars to be equipped with industry-leading compute resources to fully take advantage.

I think spending aggressively is the right approach. Meta has a right to be a leader in AI in the decades to come. That will surely benefit its existing profit drivers and developing businesses like smartglasses. 

Despite all of this minimizing the risk of overbuilding, that risk is still not zero. That’s why a quote from Zuck got me pretty excited during the call. He spoke about constant inbound requests for Meta compute renting and entertained the idea of offering that compute if they have future gluts. That would lower the risk of overbuilding and wasting dollars even more. It sounds like Meta entering the public cloud space to compete with Google, Azure and AWS is a real possibility down the road.

“Almost every week, people come to us from outside the company asking us to stand up an API service or asking if we have different compute that they could get from us. We haven't done that yet, but obviously, if you got to a point where you overbuilt, you could have that as an option.” – Founder/CEO Mark Zuckerberg

The Core Business & AI – Engagement:

The runway for AI-enabled engagement gains remains massive while the progress remains brisk.

Updates to the AI-powered content recommendation systems boosted Facebook and Threads time spent by 5% and 10%, respectively. Instagram video engagement soared by 30% Y/Y due to these changes. They deliver these types of gains every single quarter, which is how they’re still somehow growing U.S. engagement by 10% Y/Y for Facebook… their most mature market and their most mature app. Generally speaking, Facebook and Instagram engagement growth is accelerating, and it’s no secret why that’s the case. AI is not merely a future value driver for Meta. It is powering a large portion of these engagement gains today. 

While much attention was paid to the OpenAI Sora launch, I actually view AI-generated content as a large tailwind for Meta’s business. Not only do they have their own product (Vibes) in this area, but a massive injection of different content onto Meta’s apps should be fantastic for the business. It will mean models have a lot more to choose from in terms of showing users the right things, which should boost engagement. If that happens, it will mean these models have even more user data to work with, leading to more engaging experiences and incremental time spent. It's a compelling flywheel. When social media moved from friends-only content to influencer-led content, Meta greatly benefitted. Zuck thinks this transition will have a similar outcome and I agree.

Whether it’s upgrading these recommendation engines with more data, releasing better models or deploying new training and inference techniques… They have so much productive work left to do.

The Core Business & AI – Monetization:

There are a few important products to understand for this section in terms of function and connections to one another. Lattice and Gem are the foundational teaching models for advertising. They're giant generalists. They study consumer patterns and all relevant data to build an understanding of interests. They use years and years of outcome history to assign probabilities for things like clicks and conversions for a given ad. From there, retrieval models like Andromeda use this information to select an optimal batch of advertisements based on campaign objectives. Finally, runtime models rank each option and use inference to select the best ad to surface. This inference is better suited for smaller, more specialized models to avoid exploding costs. Throughout every stage of this process, there’s a boatload of progress left to enjoy.

They’ve effectively consolidated 100 models under Lattice, with 3% ad conversion gains stemming from this change and another 200 models left to combine.

They’re working hard on combining different AI systems from Instagram, Facebook and ad recommendations to make sure these platforms can openly borrow from each other and collaborate. That process keeps bearing fruit, while other areas of work like ad supply optimization are helping boost conversions even more. Retrieval models like Andromeda are infusing additional compute and data into their curated ad options, while their runtime models (picking the final ad to show) got another update that boosted Instagram conversions by 2%. So many things to do. So much improvement to deliver. So much value to enjoy as that improvement keeps coming. 

Meta AI:

Meta AI has over a billion monthly active users (MAUs) before adding its frontier models to that product. They’re confident in growth being in its early stages, with integration of products such as Vibes (AI video feed) leading to strong engagement gains with high retention. They’ve also gotten a lot better at adding creator content to Meta AI answers, which is leading to more relevant conversations and also keeping consumers on the apps for longer. New products feeding their core business heath.

Business Messaging:

Meta Business Messaging is up to 1B+ conversations per day. It’s helping companies improve and scale customer service, without ballooning costs. And it’s helping them do so right from the apps that their users are already using. Within this category, WhatsApp enjoyed 60% Y/Y growth and they’ve just added access to businesses in the Philippines and Mexico. Excitingly, it’s now letting U.S. businesses embed these tools right into their own websites, greatly expanding reach and product market fit for this segment.

Glasses:

For a review of their new smartglasses announced at Meta Connect, click here (section 2).

All we need to know for this segment is that the Meta Ray-Ban Display glasses sold out in 48 hours and has a month-long waiting list for demos. They’re working hard to increase capacity, as every launch they make in this category resonates. They’re the clear leaders in next-gen wearables.

Good quarter. I am entirely fine with Meta’s approach to spending, as I view the bountiful use cases for this compute as greatly reducing the risk of wasting sizable sums of money. This is not the Metaverse 2.0. It’s exciting to hear Zuckerberg talk about the future and easy to invest my money in that vision and him. This is not a company I worry about, although slower 2026 profit growth could lead to stock price volatility next year. Maybe... maybe not. And candidly, I don’t really care. This is a world-class company with a world-class team and world-class execution. They're dominant and extremely insulated from competitive threats. I’m going to get out of their way and let them work. My confidence in having this holding near the top of my portfolio is high and I have no interest in touching my stake at this moment. If I owned none of it today, I'd quickly build out the stake.

“Taking a step back, if we deliver even a fraction of the opportunity ahead for our existing apps and the new experiences that are possible, then I think that the next few years will be the most exciting period in our history.” – Founder/CEO Mark Zuckerberg