Meta beat earnings. Stock dropped 7%. That is not a typo, and it is not a fluke. It is the cleanest live demonstration of what Wall Street has decided AI capex is and is not allowed to look like in 2026.

What actually happened

Meta posted a Q1 beat. Revenue exceeded consensus. Earnings exceeded consensus. Reality Labs losses came in roughly where expected. Then the company raised full-year capex guidance from a $115-135 billion range to a new $125-145 billion range. That is a $10 billion increase against a Wall Street whisper number that was already considered aggressive.

Meta also disclosed that Q1 Family Daily Active People (Zuckerberg's preferred user metric) actually fell quarter-over-quarter. The company blamed "internet disruptions in Iran" tied to the ongoing war disruptions. Whether the explanation holds or not, the headline number is the same: capex up, users down, stock down 7%.

The thing the market punished

Meta is the cleanest pure-play on the most uncomfortable question in AI investing right now: where is the revenue you can put your finger on? Alphabet has a $460 billion cloud backlog. Microsoft has a $37 billion AI annualized run-rate and $190 billion in committed capex (which the market still hated, but at least it has the run-rate). Amazon has AWS at $37.6 billion in Q1 revenue alone. Meta's AI? Engagement on Reels. Better ad targeting. Llama, which it gives away. A Muse Spark model nobody has seen in production at scale yet.

There is no $460 billion backlog at Meta. There is no third-party AI revenue line. Every dollar Zuckerberg spends on H200s, B200s, GB300s, and now an entire datacenter campus in Louisiana, has to come back to shareholders through better ads, more time-on-app, and someday a glasses business that justifies Reality Labs. That is a much harder pitch in front of a Wall Street that just heard Microsoft, the most disciplined infrastructure operator in tech, admit memory prices are dragging margins.

The canary in the capex mine

Meta is the canary because it is the smallest and least-capitalized of the four hyperscalers reporting tonight. Combined 2026 hyperscaler capex now sits in the $705-725 billion range across Alphabet, Microsoft, Meta, and Amazon. Of those four, Meta has the weakest near-term path from AI spend back to AI revenue, and Wall Street knows it. The 7% punishment is what happens when investors say: we are willing to pay for AI capex that has a customer attached, and we are not willing to pay for AI capex that requires you to prove monetization through your own consumer surfaces.

If that read holds, the next leg of the AI trade is going to be brutal for any AI spender that doesn't sit between a chip vendor and a model lab or between a model lab and an enterprise customer. The middle of the value chain is where the ROI math is being underwritten. The endpoints (Nvidia, Anthropic, Microsoft Azure, Alphabet Cloud) are getting a different multiple than the consumer-AI spenders (Meta, OpenAI, even Apple to a degree).

What this means for Reality Labs

Reality Labs is no longer the most expensive bet at Meta. AI infrastructure is. That changes the political math inside the company. When Reality Labs was a $20 billion annual loss against a flat capex base, it was tolerable as a Zuckerberg moonshot. Now Reality Labs is a smaller line item than the year-over-year increase in AI server spend. The mood music inside Meta about which moonshots survive 2027 is going to shift.

Watch the May 20 layoff round. We have flagged it for weeks. 8,000 confirmed cuts, plus 6,000 cancelled open roles, all on a single day. The internal Janelle Gale memo (Bloomberg) framed it as efficiency. The 7% stock drop tonight just gave Zuckerberg cover to make those cuts more aggressive, not less.

The wider read

Tonight was the first real test of the AI capex thesis with all four hyperscalers reporting in the same window. The market said yes to the one with the contracted revenue book and the third-party customers. It said no, or at best maybe, to the three that are spending without selling the spend forward to someone else. That is the new rule, and Meta is the company most exposed to the new rule.

Sources: Meta Q1 2026 press release (April 29), CNBC Q1 earnings recap, Business Insider live blog, Yahoo Finance, Sherwood News, Ainvest cash burn analysis, BBC News market reaction, Hindustan Times big tech recap.