Facebook was cool once. Not serviceable-cool, not niche-cool — genuinely, culturally cool. When it opened beyond Harvard in 2006, people talked about it the way they talked about getting into a good restaurant. The invitation carried weight.
That era is so far gone it’s almost archaeological.
The Numbers Nobody Wanted to Admit
On April 29, Meta reported something it had managed to avoid for seven years: fewer daily active users. The total across its “Family” of apps — Facebook, Instagram, WhatsApp, Messenger — dropped by 20 million, landing at 3.56 billion for Q1 2026. Zuckerberg blamed internet disruptions in Iran and a WhatsApp restriction in Russia. Analysts, predictably, weren’t buying it.
The revenue story looks different on the surface. Ad revenue hit $56.31 billion, up 33% year over year. The stock still dropped over 7% after earnings. That gap — between a company printing money and a market that’s spooked — tells you everything. Wall Street isn’t pricing Meta on what it earns today. It’s pricing the question of whether the engine that generates those earnings has a future.
Investigative journalist Julia Angwin made the structural case in the New York Times shortly after the earnings drop: this is a company entering its decline phase, the long, slow rot that took Yahoo and AOL from dominant to irrelevant. The parallel isn’t flattering, but it’s hard to argue with.
The AI Bet That Keeps Not Paying Off
Meta has spent — and continues to spend — at a scale that’s hard to fully comprehend. Capital expenditures for 2026 are projected at $125–145 billion, up another $10 billion from prior estimates, almost all of it funneled into AI infrastructure. The company’s Superintelligence Labs, built around Alexandr Wang after Yann LeCun’s messy departure, recently released Muse Spark, its first closed-source model. An executive told Bloomberg it won’t keep up with ChatGPT, Claude, or Gemini. The company’s own blog described it as “an early data point on our trajectory.”
That’s a polite way of saying: we’re behind.
The Llama 4 debacle still lingers. LeCun himself told the Financial Times that benchmark results “were fudged a little bit” — and that Zuckerberg “basically lost confidence in everyone involved.” The ensuing talent exodus and expensive rebuilding campaign bought Meta a model that, by its own admission, can’t compete. The Reality Labs division, still nursing the Metaverse wound, posted another $4.03 billion operating loss for the quarter.
What’s Actually Happening to the Feeds
The product that still generates all that revenue — Facebook — has become, in 2026, something genuinely difficult to describe to someone who wasn’t around for 2008. What was once a way to keep up with people you actually knew is now an algorithmic content churn of AI-generated slop, engagement bait, and ads that know too much about you. Instagram isn’t far behind. The company’s plan to prioritize “original content” on both platforms, announced alongside Q1 earnings, reads as an acknowledgment that they’ve let the quality deteriorate past a sustainable floor.
Meta has 8,000 layoffs coming. Ten percent of the workforce. The company is simultaneously slashing headcount and dramatically accelerating capital spending — a pattern that makes sense only if you believe the current product portfolio is a temporary problem, and AI is the permanent solution.
The Extraction Problem
There’s a structural argument that explains the user decline better than any single product failure. Meta’s original value proposition was connection — a genuine utility that earned trust and loyalty over time. What the company optimized into, over the course of a decade, is a data and attention extraction engine. The goodwill reserve that once let it weather scandal after scandal has quietly emptied. Legal systems are starting to reflect this reality — a New Mexico jury ordering $375 million for failure to protect children is not a one-off fine, it’s a signal that the company’s approach to harm is no longer being treated as incidental.
The money still flows. The users are still there, 3.56 billion of them. But the first crack in the user metric, after seven years of unbroken growth, is the kind of thing that tends not to reverse on its own.
What Zuckerberg needs isn’t a better AI model. It’s a reason for people to care again. And that’s the one thing $145 billion in capex can’t buy.
Related: China Just Rewrote the Rules of AI M&A — and Meta Is Paying the Price
