Meta just delivered its biggest year-over-year revenue jump since the pandemic. A solid 33% growth to $56.3 billion, ad impressions up 19%, ad pricing up 12%, and operating margin holding firm at 41%. By any reasonable measure, Q1 2026 was a great quarter.
The company raised its 2026 capex guidance to $125-145 billion, up from a prior range of $115-135 billion. Combined with a $107 billion step-up in contractual commitments disclosed during the quarter and the announcement of a layoff coming in May, the market is worried… Zuck is going long on a vision again, and shareholders should brace.
Is this… going to be another metaverse?
In short: nope, I don’t think so. But the full answer is a little bit more nuanced. Let me explain.
Meta’s AI ambitions split into three buckets, distinguished by how close they are to revenue, how measurable the returns are, and how long the runway is to monetization.
Bucket A: the engine that’s already working
This is core machine learning work. The models that decide which Reel you see next, which ad gets shown after that Reel, and which version of which ad creative gets selected from the candidate pool.
The cleanest evidence that Bucket A is paying off is in the impression numbers. In Q1 2026, ad impressions across Meta’s family of apps grew 19% year-over-year while Daily Active People grew only about 4%.
This means that roughly fifteen percentage points of impression growth came from people spending more time in the apps, and from Meta showing more relevant ads in the time they were already spending. Both are products of the AI ranking and recommendation work.
Impression growth is only half of it. Average price per ad grew 12% year-over-year in Q1 2026 — an acceleration from 6% in Q4 2025. Both impressions and pricing growing at double-digit rates simultaneously is the strongest possible signal that Bucket A is working.
It means the ads themselves are getting more valuable to advertisers. That is exactly what you’d expect when AI targeting improves, conversion rates lift, and advertisers respond by bidding higher because their return on ad spend is better.
The mechanics behind these gains are worth understanding because they explain why the gap is likely to persist as they continue to invest in AI. Meta has been steadily upgrading the layers of its ads pipeline. Andromeda narrows the candidate pool from tens of millions of potential ads down to a few thousand. GEM, the Generative Ads Recommendation Model, is the largest training-side architecture, and its intelligence gets distilled into a smaller, faster model — the Adaptive Ranking Model — that serves ads in milliseconds. Lattice is the underlying architecture that consolidates dozens of older specialized models into fewer, more capable ones.
In Q1 2026 alone, enhancements to Lattice and GEM drove more than 6% conversion rate improvement on landing-page-view ads, and the Adaptive Ranking Model expansion drove an additional 1.6% conversion lift across major Facebook and Instagram surfaces.
And these gains compound. Better recommendations make people spend more time in the apps. More time spent generates more ad inventory at constant ad load. More ad inventory generates more impressions. More impressions generate more training data. More training data improves the next generation of models. The flywheel turns, and each turn enhances Meta’s moat.
This is the bucket where Meta’s AI investment is unambiguously earning a return today. And the seeds were planted quarters ago, CFO Susan Li was explicit about this on the Q2 2025 call:
“On the core AI side, we continue to see strong ROI. Our ability to measure that is quite good, and we feel sort of very good about the rigorous measurement and returns that we see there.”
Bucket B: the probable next wave?
These are products and services that have users and (maybe) revenue already… Meta AI, the Ray-Ban Meta and Oakley AI glasses, and Business AIs running on WhatsApp and Messenger… but aren’t moving the needle yet.
Glasses. Reality Labs revenue was $402 million in Q1 2026, down 2% year-over-year as Quest headset declines partially offset growing AI glasses revenue. Daily active users on AI glasses tripled year-over-year. Zuckerberg called the category “one of the fastest-growing categories of consumer electronics ever” on the earnings call.
But hardware sales alone are not the real prize. Zuckerberg has been explicit for over a year that he sees glasses as the next computing platform after the smartphone, telling investors on the Q4 2024 call that this is “the next computing platform” Meta has been building toward. He wants to own the platform layer the way Apple owns iOS: the operating system, the AI assistant, the ad surface, the ecosystem.
Business messaging. Family of Apps “Other revenue” hit $885 million in Q1 2026, up 74% year-over-year, driven primarily by WhatsApp paid messaging and subscriptions. What hasn’t been switched on is the agent layer on top: over ten million weekly conversations now run through Business AIs, up from one million at the start of this year. CFO Susan Li was direct on the Q1 2026 call: “Business AIs today are currently free for most businesses on our messaging apps.” The agent monetization model is not yet deployed.
Meta AI. The standalone consumer assistant is free, with no ads or premium tier. After Muse Spark replaced the prior model, Meta reported double-digit percent increases in sessions per user. Zuckerberg has said for over a year that monetization comes after scaling — the same playbook Meta has run for every social product in its history.
Bucket C: the leap of faith
This bucket started with much fanfare where Zuck recruited elite AI talent at extraordinary cost, which gave rise to the Meta Superintelligence Labs (MSL).
Its first model, Muse Spark, has shipped and now powers Meta AI. Beyond that, MSL is training successor models, building out training clusters that will scale to multiple gigawatts of compute, and continuing to recruit elite AI researchers at extraordinary cost.
There’s currently no directly measurable returns from this bucket yet, but Zuck made an argument for why this is important on the Q1 2026 earnings call:
“You can’t build a leading AI product if you don’t have leading models. So you’re not going to have leading models in the future if your models can’t improve themselves… If we don’t have an ability to do that, then we or anyone else, the companies that don’t do that are not going to be leading labs, they’re not going to produce leading products.”
In other words, Zuck believes that having a leading frontier model will have a trickle-down effect to other products.
When pressed for an answer by Morgan Stanley’s analyst during the earnings call on what specific signposts he was watching to ensure healthy ROIC on all the AI infrastructure spend over the next 12 to 24 months. Zuckerberg pivoted away from the ROIC frame entirely, offering instead a qualitative three-step playbook — “first, technically, are we delivering the quality to enable a great product; then second, when you have the product, how is it scaling; and then third, you look at the monetization” — and admitting plainly: “I don’t think we have a very precise plan for exactly how each product is going to scale month-over-month.“
That’s about as honest as it gets during an earnings call and probably contributed to why the stock sold off after the earnings. Zuck is basically saying… #trustmebro
So where is the CapEx actually going?
Zuck is investing in AI like this is an existential risk to Meta’s business. And he probably should be. Meta’s entire business depends on users showing up to their apps every day, and if AI represents a generational shift in how people consume information and entertainment, how creators create content, or how advertisers get the most bang for their bucks, Meta has to be at the frontier or it loses relevance.
But I thought it’s worth pointing out the structural difference between how Meta is investing and how the hyperscalers (Microsoft, Amazon, Google) are investing.
The hyperscalers are building compute to rent out (Google is a mixed bag). Their ROI depends on enterprise customers — including frontier labs like OpenAI and Anthropic — paying for that capacity at a price that justifies the build. If AI demand softens or capacity outpaces demand, the hyperscalers may face oversupply and pricing pressure. Their CapEx bet is fundamentally on someone else’s willingness to pay.
Meta isn’t building to rent out. Meta is building for its own consumption, primarily for ads ranking, content recommendations, and increasingly for inference on consumer AI products.
The “customer” is Meta itself. And the demand signal is already visible in the numbers: 19% impression growth, 12% pricing growth, and the conversion rate lifts from each model upgrade. Susan Li and Mark Zuckerberg have both said repeatedly that they have been underestimating how much compute is needed.
Meta has a clearer line of sight into the demand, because they generate it themselves. And the existing results validate that returns on this investment are real.
The risk in this framing is that Meta’s compute demand depends on users continuing to show up to their apps. If user attention erodes — to TikTok, to ChatGPT-style assistants, to whatever comes next — the demand thesis weakens. That’s the actual bear case, and it’s a real one. But Q1 2026 doesn’t show that erosion happening. If anything, the AI investments are deepening engagement (more time spent on Reels, better content recommendations) rather than the reverse.
Even if Bucket C disappoints, the infrastructure is fungible. CFO Susan Li was direct on the Q1 2026 call:
“So we’re going to continue building out our infrastructure with flexibility in mind. And if we end up not needing as much as we anticipate, we can choose to bring it online more slowly or reduce our spending in future years as we grow into the capacity that we’re building now.”
The combination — line of sight on demand, fungibility of capacity, willingness to flex — means this CapEx looks much less binary than the headline number suggests.
The bottom line
Meta continues to trade at a very undemanding forward PE multiple of 18.6x. The company is basically firing on all cylinders, but the market continues to be worried about two things…
Firstly, that Mark Zuckerberg is going into his “Metaverse” mode with AI. But based on whatever has happened, with the accelerating revenue growth, strong impressions and ad prices growth, especially at a time when other companies are blaming macro, I’d say we’re getting good returns from their earlier investments in AI, and I much rather they invest ahead than to rest on their laurels. Time will tell if this venture turns out well, and if it doesn’t pan out, whether CFO Susan Li is able to steer Zuck into the right capital allocation decisions.
Secondly, I think there’ll always be a dark cloud that lingers over Meta every now and then. Perhaps because this isn’t an easy business to run. A social media company that crushes it out of the park for so long is the exception, not the rule, and Zuck has done a phenomenal job. No other social media company is anywhere near Meta.
And as long as Zuck is at the helm… and I’ve said this for a couple years now, I’m going to let him cook.
Disclaimer: These research reports constitute the author’s personal views and are for educational purposes only. It is not to be construed as financial advice in any shape or form. From time to time, the author may hold positions in the stocks mentioned below that are consistent with the views and opinions expressed in this article. Disclosure – I have a position at Meta at the time of publishing this article (this is a disclosure and NOT A RECOMMENDATION).