There’s a specific category of earnings report that I call the “technically great but emotionally catastrophic” quarter. Revenue beats estimates. EPS comes in above expectations. User growth holds. Operating income rises. And then somewhere in the prepared remarks, management mentions a number that causes institutional investors to put down their coffee cups and stare at the ceiling.
For Meta’s Q1 2026, that number was $125 to $145 billion.
That’s the updated capital expenditure range for the full year, raised from the prior guidance of $115 to $135 billion. To be clear about what $145 billion means: it’s more than the entire annual GDP of Hungary. It’s more than Warren Buffett’s entire cash pile before he deployed it. Meta is spending it on data centers, custom silicon, and the AI infrastructure that Mark Zuckerberg has decided represents the company’s future — whether investors like the timeline or not.
The stock dropped 10% the morning after results. Which is its own kind of statement about where investor psychology sits on the question of AI spending right now.
What Actually Happened in Q1
Revenue came in at $56.31 billion, up 33% year over year. That’s the fastest growth rate Meta has posted since 2021, which was itself a year when ad prices were running hot because the world had emerged from COVID and every company with a marketing budget was spending it online. The growth in 2026 is different — it’s not a post-lockdown sugar rush, it’s AI-driven advertising efficiency actually showing up in the numbers.
Operating income hit $22.9 billion, up roughly 30% year over year. EPS came in at $7.31 excluding a one-time $8 billion tax benefit — well above what analysts had modeled. For Q2, the company guided revenue between $58 and $61 billion, which, if the midpoint lands, would represent another quarter of roughly 30% growth.
None of those numbers explain a 10% stock drop. The capex does.
Total 2026 expenses are expected to land between $162 and $169 billion. That figure includes the $125 to $145 billion in capex — a number that, compared to the roughly $72 billion Meta spent on capex in 2025, represents a doubling of infrastructure investment in a single year. The company attributed the increase to “higher component pricing and additional data center costs to support future year capacity.” Which is corporate-speak for: NVIDIA chips got more expensive and we’re building more stuff than we originally planned.
| Meta Q1 2026 Key Metrics | Result | vs. Year Ago | vs. Estimate |
|---|---|---|---|
| Total Revenue | $56.31B | +33% | Beat |
| Operating Income | $22.9B | +30% | Beat |
| EPS (ex. tax benefit) | $7.31 | significant increase | Beat |
| Capex (Q1 alone) | $19.84B | — | Below $27.57B est. |
| Full-Year Capex Guide | $125-145B | vs. $72B in 2025 | Raised significantly |
| Q2 Revenue Guide | $58-61B | — | Above estimates |
The interesting wrinkle in the capex story is that Q1 capex actually came in below analyst estimates — $19.84 billion versus an expected $27.57 billion. Meta underspent in the quarter but raised the full-year guide. What that tells you is that the company is backloading the spending, not pulling back from it. The market reacted to the full-year guide, not the quarterly number.
The Advertising Business Nobody Wants to Talk About
Everyone wants to discuss Llama and superintelligence and Zuckerberg’s long game. Fine. We’ll get there. But first, can we acknowledge that Meta’s core advertising business is doing something genuinely unusual?
Thirty-three percent revenue growth with no major new product launches. That’s not normal. Facebook has existed for 20 years. Instagram has been around since 2010. WhatsApp celebrated its 17th birthday. These are not products in early growth phases — they are mature platforms that most analysts two years ago expected to generate mid-single-digit revenue growth at best.
The reason they’re growing at 33% is that AI-powered ad targeting has materially improved advertiser return on investment, which means advertisers are willing to pay more per impression because the impressions are actually working. When your ad system can identify with more precision which users are likely to convert on a given product — not based on demographic categories but based on behavioral signals across a family of apps used by 3.4 billion people daily — you have built what is effectively the most efficient advertising machine in human history.
This matters enormously for the investment case because it means the AI spending isn’t purely speculative. Some of it is already paying back through the ad revenue line. The question is whether the next phase of the spending — the foundational model development, the AI assistant rollout across apps, the enterprise tools that don’t exist yet — generates comparable returns on comparable timelines.
The Llama Strategy: Genuinely Clever or Genuinely Reckless
Llama has been downloaded over 1.2 billion times. If you know anything about open-source software, that number stops you cold.
For context, Linux — the operating system that runs most of the internet’s servers, most cloud infrastructure, and most Android phones — took decades to achieve that kind of adoption. Llama, Meta’s open-source large language model, has moved from research project to foundational infrastructure for a significant portion of the global AI developer ecosystem in roughly two years.
The strategy behind this is more deliberate than it looks from the outside. By open-sourcing Llama, Meta has made itself the default starting point for developers building AI applications — not because Meta requires it, but because starting with a capable open-source model and customizing it is faster and cheaper than building from scratch. Every developer who builds on Llama learns the Meta ecosystem. Every application built on Llama generates implicit information about what users want from AI. And every enterprise that builds internal tools on Llama is one that Meta can potentially sell enterprise AI services to later, once that monetization layer exists.
Compare this to Google’s approach with Gemini, which is proprietary and monetized directly, or to OpenAI’s API model. Meta is playing a different game — ecosystem capture before monetization, in the pattern of how Amazon gave away AWS cheaply to build cloud dominance before raising prices, or how Android gave away the mobile operating system to capture distribution.
The risk is timing. Ecosystem capture strategies work over five to seven year horizons, not two to three. The $125 to $145 billion in 2026 capex is being spent on a strategy whose payoff is probably 2028 or 2029 at the earliest. Investors pricing the stock on next year’s earnings multiples don’t find that answer satisfying.
The Reality Labs Elephant in the Room
I have to mention it even though nobody wants to.
Meta’s Reality Labs division — the augmented and virtual reality unit — lost $4.97 billion in Q1 2026 alone. That’s roughly $20 billion annualized. The cumulative losses in Reality Labs since 2020 now exceed $60 billion. The Ray-Ban Meta smart glasses have been the most commercially successful product from this division, which, given the context of those losses, is damning with genuinely faint praise.
Zuckerberg has not abandoned the vision. He probably won’t. He controls the company through a dual-class share structure that means shareholders can express displeasure but cannot force a change in strategy. The losses continue.
I’ve made my peace with Reality Labs as a permanent drag on Meta’s otherwise extraordinary business. Some investors haven’t. If you’re buying Meta, you’re buying the ad business and the AI bet, with a $20 billion annual tax attached for Zuckerberg’s metaverse conviction. Whether that’s acceptable is a personal decision that valuation cannot settle.
The Regulatory Situation, Which Is Always Fine Until It Isn’t
EU probes. UK Online Safety Act disputes. A blocked acquisition in China. An ongoing antitrust case in the United States that has been working its way through the courts with the energy and pace of continental drift.
Meta has been living under regulatory pressure for the better part of a decade and its revenue has roughly quintupled in that time, which probably tells you something about the practical impact of regulatory scrutiny on the actual business. The more serious risk isn’t fines — Meta can absorb fines — it’s forced structural changes. A forced separation of Instagram from Facebook would be a genuine problem. The probability of that happening has been debated by antitrust lawyers since 2020. It has not happened yet.
I’m not dismissing regulatory risk entirely. The FTC case in particular has dragged on long enough that its eventual resolution matters. But I also know that markets have priced in regulatory catastrophe for Meta roughly a dozen times in the last eight years, and the stock kept going up despite it. At some point the “regulatory overhang” becomes background noise that sophisticated investors discount appropriately.
Valuation: The Part Where the Humor Goes Quiet
Meta is currently trading around $613. That’s down roughly 23% from the $796 high it hit earlier this year. The forward P/E sits around 24 times consensus 2026 earnings.
That multiple, in isolation, looks reasonable. For a company growing revenue at 33%, a 24x forward P/E would normally be considered a good deal. The complication is that the current earnings base is going to face meaningful pressure from the capex ramp. You are not buying 2026 earnings at 24x — you are buying earnings that will likely be lower than their current run rate in 2027 because the company is reinvesting at a rate that will compress near-term margins before expanding them.
The more interesting frame is to think about what the business looks like in 2028 if the AI spending actually works. If Llama-based enterprise products generate even $20 to $30 billion in annual revenue by then, if ad revenue keeps compounding at something approaching current rates, and if Reality Labs losses eventually stabilize — which is admittedly the most optimistic of the three assumptions — Meta’s earnings power in 2028 looks substantially different from today’s.
The analyst consensus agrees with this framing. 96% of Wall Street analysts rate Meta a Buy. Average price target is somewhere around $840, implying roughly 37% upside from current prices. These are not people who are confused about the capex; they’ve modeled it. They’ve concluded that the business quality, the ad engine’s continued growth, and the option value of the AI ecosystem justify owning the stock through the spending cycle.
The bear case is simpler and less fun to argue because it’s basically: what if the AI stuff doesn’t work on any reasonable timeline and Zuckerberg keeps spending $125 billion a year anyway. That scenario produces genuine pain. It’s also the scenario where every single metric in the core business stays strong while the stock underperforms, which is a specific kind of frustrating that doesn’t get discussed often enough.
Where I Actually Land
The $613 price and the 23% pullback from the high have made this more interesting than it was six months ago. I’m not going to pretend the valuation is screaming obvious — it isn’t. But I’ve been thinking about this company for three years, and a few things keep bringing me back.
The advertising business is fundamentally a different quality asset than it was in 2022. The AI improvements to targeting have created a flywheel where better targeting produces higher advertiser ROI, which produces more advertiser spend, which funds more AI development, which produces better targeting. That loop is running and it doesn’t require any speculative future product to explain the current revenue growth.
The Llama ecosystem play is a real long-term bet with real logic behind it. It’s not guaranteed. But it’s not delusional either — it’s the kind of platform strategy that has worked before in adjacent contexts.
And the valuation, at 24x forward earnings for a business growing revenue at 33% with operating margins above 40%, is not unreasonable by most frameworks I use. The capex creates near-term earnings uncertainty. It doesn’t change the quality of the underlying asset.
I’d own this here. I’d hold more of it if the market gives us another leg down on capex anxiety. And every quarter I’d be watching the ad revenue growth rate closely — because the moment that starts decelerating toward 15% or below, the AI spending no longer has the same cover it does when the core engine is running this hot.