Meta's Q3: $64.6B Revenue, a Stock Split Built for AI Capital, and $825B in Commitments

Meta’s Q3 earnings, reported October 21, contained the usual headline and an unusual corporate structure change. The headline: revenue of $64.6 billion, up 26% year over year, with ad revenue at $63.06 billion on 10% more ad impressions and 3.64 billion daily active people, per the earnings coverage from Bloomberg, WSJ, and Reuters. The structure change: a 2-for-1 stock split executed through new nonvoting Class C shares, one for each existing Class A share, with trading expected around January 2027, per WSJ’s report on the filing. Zuckerberg’s voting control, roughly 60.6%, is untouched; what the split actually creates is capacity to issue equity for AI spending without diluting the founder’s vote, which is why one analysis called it “a tax deal dressed as a stock split.”

The AI bill, in three numbers

Three figures from the quarter define Meta’s AI position. First, 2026 capex guidance narrowed to $135-145 billion, with Q3 alone at $34.91 billion. Second, per Bloomberg’s October 22 report on the filings, Meta’s lease and contract commitments reached $825 billion, up from roughly $700 billion in July, including $355.7 billion of off-balance-sheet leases through 2036; that is data center and compute capacity contracted years into the future, now larger than most countries’ infrastructure budgets. Third, the quarter absorbed a $9.89 billion legal settlement charge that pushed operating income down 37% despite the revenue growth, a reminder that even at Meta’s scale, one-time items can consume a third of operating profit. Against this, the contrast with last week’s Microsoft-Nscale cancellation and the UK’s grid-constrained projects is stark: Meta is behaving as if power and capital are solvable, and its balance sheet is the proof of commitment.

Muse is the consumer proof point

The earnings call supplied the first hard numbers for Meta’s consumer agent: 10 million US downloads in 40 days, per Sensor Tower data via TechCrunch, making Muse only the twelfth app ever to reach that pace, with retention running ahead of AI-app averages. Zuckerberg claimed more than 25% of Muse users make purchases averaging over $200, and Gene Munster estimated roughly 10 million daily users with a path to a billion. Skepticism is warranted on the commerce conversion figures, which come from the company selling the story, but the download trajectory is independently measured and it explains the strategic repositioning: Muse Spark Open, the model behind the assistant, was open-sourced six days before earnings, and Meta entered what Engadget called “a new phase” by treating the agent as a commerce surface rather than a chatbot. Anthropic’s answer, Claude Duet, launched the same week on the opposite thesis: fewer users, deeper trust.

What to watch

Three things matter beyond the quarter. First, whether the Class C issuance actually prices: a nonvoting share class trades at a discount, and the size of that discount is the market’s price tag on Zuckerberg’s control; watch the registration statement for the planned raise amount. Second, whether the $825 billion commitment figure keeps climbing faster than revenue; the gap between 26% revenue growth and commitments that jumped $125 billion in a quarter is the entire AI-capex debate in two numbers, and Meta’s Q4 guidance of $73-76 billion in revenue suggests the company is betting inference demand, especially from Muse and business agents, closes it. Third, Europe: Wired reported EU ad growth slowed to 18% year over year after the Less Personalized Ads rollout, and if that trend steepens, the ad engine funding the AI buildout gets a second structural headwind alongside the capex bill. Meta is making the largest private infrastructure bet in corporate history on the assumption that agents become the next ad surface; this quarter was the first one where the consumer data supported the assumption.