Mark Zuckerberg is spending $65 billion a year on AI infrastructure and Wall Street is treating it like a midlife crisis. The stock is down 25% from its highs. The analyst notes read like concerned letters from a parent. "Too much capex." "When does it pay back?" "Where's the ROI?"
Meanwhile, Meta just dropped a quarter with $56.3 billion in revenue, $26.8 billion in net income, and operating margins north of 40%. That's more profit in 90 days than most Fortune 500 companies make in a year. And the growth rate isn't slowing — it's accelerating.
So here's the question everyone should be asking but isn't: what if this isn't a spending problem — what if it's a pricing problem? Because right now you can buy Meta at 15 times forward earnings. For a company growing revenue 33% year-over-year, with 3.3 billion people using its apps every single day. Coca-Cola trades at 22 times earnings and grows at 5%. The math on Meta isn't just attractive. It's broken.
The AI infrastructure story isn't theoretical anymore. Meta's ad platform now runs on a foundation of proprietary AI models — Llama, Andromeda, and a recommendation engine that's been rebuilt from the ground up on custom silicon. Every ad you see on Instagram, every Reel that keeps you scrolling, every product listing in Facebook Marketplace is the output of models Meta trained on its own infrastructure. The capex isn't a bet. It's the factory floor.
Here's what $65 billion in annual capex actually buys. It buys data centers the size of shopping malls running on custom MTIA chips instead of paying NVIDIA's markup. It buys the Llama 4 model family that's already powering AI features across Facebook, Instagram, WhatsApp, and Messenger. It buys the compute to train next-generation recommendation systems that don't just match ads to users — they predict what you'll want to buy before you know it yourself.
And here's what it's already producing. Ad revenue grew 33% year-over-year. That's not a typo. The ad business that was supposedly getting disrupted by Apple's privacy changes and TikTok's rise just posted one of its best growth rates since the pandemic boom. Ad impressions are up. Ad prices are up. And AI-powered tools like Advantage+ shopping campaigns are seeing conversion rates that make Google's search ads look pedestrian.
The market keeps pricing Meta like a social media company that's burning cash on a science project. But the numbers tell a different story. Gross margins sit at 81.9%. That means for every dollar of revenue, Meta keeps nearly 82 cents before operating costs. Operating margins are 40.6%. Return on equity is 32.9%. Free cash flow hit $25.6 billion — in a year where capex was supposedly "out of control." The company is literally spending tens of billions on infrastructure and still printing more cash than it knows what to do with.
There's a legitimate bear case here, and it's worth addressing. Reality Labs — the metaverse division — burned through roughly $17 billion last year and generated pocket change in revenue. That's real. The regulatory risk in Europe and the U.S. around AI models and data privacy is genuine. Competition from TikTok isn't going away. But here's the thing: even if you value Reality Labs at exactly zero dollars, the core ad business alone is worth north of $1.3 trillion on a 20x multiple of its earnings. The entire company trades at $1.4 trillion. You're getting a moonshot AI lab and a metaverse division for free.
The analyst community isn't conflicted about this. Fifty-nine analysts cover Meta and the average rating is a Strong Buy. The mean price target sits at $827 — about 50% above where the stock trades today. The high target is $1,015. These aren't meme stock analysts from a Discord server. These are institutional desks at Goldman, Morgan Stanley, and JPMorgan who've been covering this name since the IPO.
Institutional holders own 79.3% of the float. Short interest is just 1.4%. The smart money isn't betting against this. It's waiting.
Every generation of tech investing has a moment where the market panics about capex right before the returns become undeniable. Amazon spending billions on AWS while Wall Street screamed about margins. Microsoft pouring money into Azure while analysts called it a distraction from Office. Google building data centers nobody thought would pay back. In every single case, the infrastructure spend became the moat.
Meta is building the infrastructure layer for AI-powered advertising across the world's largest social graph. Three point three billion monthly active users. A family of apps that reaches more people than any other product in human history. And a recommendation engine that gets smarter with every dollar of infrastructure it sits on. You can panic about the capex. Or you can look at the forward P/E, look at the growth rate, and realize the market is handing you a gift.
Disclosure: The Signal holds no position in META. This is not financial advice.





