Meta earnings: record $64.6B revenue, Muse's million-dollar shopping flow, and a C-share split to fund the AI race

Meta delivered a report that flipped the market from punishing its spending to rewarding its execution: Q3 revenue hit a record $64.6 billion (+26% year-on-year, beating the $63.3B consensus) and Q4 guidance of $73–76 billion also topped expectations at the midpoint; a one-time $9.89 billion legal settlement crushed GAAP EPS to $3.83, but ex-item EPS was $7.40 — and the stock jumped as much as 6% after hours^[1][2].

The scoreboard: record revenue, negative FCF, higher capex

By segment: ad impressions grew 10% with average ad prices up 14%, and family daily actives reached 3.64 billion (+3%) — the ad engine remains strong^[1]. On capex, Q3 spending including finance leases hit $34.91 billion, and the company raised the floor of its 2026 capex guidance from $130 billion to $135 billion (ceiling unchanged at $145B); free cash flow was -$2.2 billion — negative for the first time since listing, but far better than the -$9.4 billion analysts expected^[1]. The EPS needs two readings: GAAP $3.83 missed the ~$6.54 consensus, but that was the $9.89 billion legal settlement; ex-item, $7.40 beat comfortably — and the market clearly chose the second reading^[1].

The call's star: Muse's ten million users and "forced-earlier" monetization

Most of the hour-long call went to three questions: how do agents make money, how is compute cost covered, and where does the money come from. Management disclosed that more than 10 million people in the US and Canada alone have set up a Muse personal agent, about 25% of them using it on any given day; Zuckerberg said shopping is Muse's most common use case — over a quarter of Muse users shop with it, average purchase above $200, with first-week volumes so far above expectations that "we had to work with Stripe to raise credit limits"^[2]. The monetization framework got its first clear articulation: free for most users, merchants pay in transaction scenarios, personal subscriptions unlock stronger models, enterprises pay for extra features. Pressed by Evercore on whether Meta needs to monetize earlier than its usual playbook, Zuckerberg broke with tradition: "I think we'll push commercialization earlier than our normal cadence — the scale of costs and opportunity is large, users are already shopping with Muse, and the payment model needs to be built alongside"^[2]. TechCrunch confirms the scale: Muse has crossed 10 million users with shopping the top activity — roughly six weeks after launch^[3].

The capital-structure bombshell: a C-share "split" and Zuckerberg's California pledge

The heaviest announcement was the creation of non-voting Class C common stock, effective January 2027, distributed as a stock dividend (one new share per share held — effectively a 2-for-1 split)^[1]. The arrangement kills three birds: it opens an equity-funding channel for AI investment — against $135–145 billion of annual capex, special-committee minutes hint that credit conditions for mega-cap tech are tightening and mega debt issuance could hurt ratings, making equity the inevitable route; it dilutes no one's voting power; and it locks in Zuckerberg — in exchange, he committed to remain in California or Meta's headquarters jurisdiction for eight years, defusing the risk that California's proposed billionaire wealth tax would force him to sell voting B shares (he may immediately sell 20% of the C shares received, about 10% of his total economic interest)^[1]. CFO Susan Li's framing was blunt: the C shares are primarily a financing instrument^[1]. Zuckerberg went further on the call: Muse's cost efficiency will improve tenfold next year, and Meta is "entering a new phase"^[2].

What to watch

First, Muse's global rollout pace — compute hunger is the binding constraint, and the "tenfold" cost-efficiency promise is the key test. Second, the first actual financing move after the C shares land — its size will set the tone for the AI arms race's ammunition depot. Third, follow-through on the legal settlement. Fourth, the rest of this week's tech earnings — Netflix has already flagged an ad slowdown, and Meta's ad engine is the industry's last ballast^[1][2].

For the market, the report's real significance is moving the "AI spend versus return" debate from the financial statements onto product data: with 10 million users, 25% daily activity and a $200 basket on the table, AI spending has, for the first time, a directly computable unit-economics model — the real reason after-hours money was willing to reprice Meta.