Earnings Preview | Meta (META.US) Faces the "AI Monetization Exam": Can Muse's Huge Success Ease Concerns Over "Burning Cash"?

According to Zhitong Finance APP, Meta (META.US) will release its 2026 Q3 financial results after the US market closes on October 21st. The market has revalued Meta from an “AI spender” to an “AI monetizer”, and this financial report needs to prove whether this transformation is real. Since the July earnings report, Meta has risen about 37%, significantly outperforming the Nasdaq 100’s 6.7% and the S&P 500’s 4.4%—though during the same period, analysts have lowered earnings forecasts by about 4%. After the recent surge, valuations appear to have fully priced in optimistic expectations.

Three Key Numbers for Q3

First is growth rate. The official guidance for revenue is $61–64 billion, up 19%–25% from $51.2 billion a year ago; the midpoint is about 22%—still high growth, but below Q2’s 28%. Analyst median estimates expect revenue of $63.29 billion, adjusted EPS of $6.85, operating profit of $18.53 billion, EBITDA of $26.99 billion, and a gross margin of 80.16%. The implied operating margin for Q3 is about 29.3%, down from 40.1% a year earlier. Market expectations for Q4 revenue are $73.56 billion, so guidance for next quarter is even more important than a Q3 “beat.”

Second is the legal accrual. On August 26, Meta reached a settlement with various state attorneys general regarding youth safety, agreeing to pay about $18 billion over ten years (about $5.3 billion of which depends on whether TikTok and YouTube take similar measures). Investors breathed a sigh of relief over the clearing of major legal risks, boosting the stock, but this expense wasn’t included in July’s spend forecast. Meta expects Q3 legal expenses to be about $10 billion, which, given approximately 2.57 billion shares, translates to about $3.9 pre-tax per share—GAAP EPS will look ugly, though the market is expecting this.

Third is 2027 capital expenditures. This is the item most likely to trigger a sharp stock price reaction. Management refused to provide a 2027 outlook in the Q2 earnings call, and shares fell about 10% the next day. Sell-side estimates diverge dramatically: FactSet consensus is ~$193.88 billion; Michael Morris of Guggenheim raised his estimate from $203 billion to $225 billion; Youssef Squali of Truist raised his estimate from $180.11 billion to $190 billion (up 32% y/y). Morris believes Meta will continue to aggressively expand its AI infrastructure, “We expect free cash flow will remain under pressure through 2028, returning to positive growth in 2029.” Squali put it more bluntly: “Whether large-scale AI infrastructure investment can convert into actual paid service revenue is critical for further stock appreciation.”

Key Factors

The advertising engine remains the ace, but the cost side is already alarming

Q2 was one of the best quarters ever for Meta’s ad business: revenue $60.8 billion, +28% y/y (27% at constant currency), ad impressions up 14%, average ad price up 12%—growth in both volume and price; Family daily actives 3.6 billion, +3% y/y; Family of Apps (FoA) operating profit $23.4 billion, while Reality Labs lost $4.6 billion.

AI’s contribution is now verifiable: Instagram usage time showed double-digit growth, Advantage+ advert solutions reached annualized revenue of over $75 billion, and Facebook ad conversion rates improved 15.7%. This is the main argument for bulls—AI improves the efficiency of ads, and advertisers are willing to pay more as a result.

But costs have shifted dramatically in the same report. Total costs rose 55% y/y to $42 billion, including $2.4 billion in legal expenses and $1.2 billion in severance (linked to about 8,000 layoffs in May 2026). As a result, operating profit fell 8% to $18.8 billion, operating margin dropped from 43% to 31%, and diluted EPS fell 13% to $6.18.

Muse: Outstanding subscription performance

Muse was the direct catalyst for this wave of gains. On September 8, Meta launched this personal AI agent, pushing the stock up about 27% during September; last week they launched the new model Muse Wonder, which features improved memory, focus, and complex task handling—stock gained another 2%+ over five trading days.

App download numbers are indeed impressive, though sources differ widely: Sensor Tower estimates Muse reached 5 million downloads in the US and Canada within 22 days, “much faster than ChatGPT”; another report puts downloads near 9 million; another claims 6.6 million in 16 days. Another Sensor Tower figure puts Muse at 2.8 million downloads globally in its first two weeks, versus ChatGPT at 3.7 million—in this estimate, Muse did not outpace ChatGPT.

More importantly, retention: Sensor Tower data shows Muse’s 14-day retention at 22%, slightly below ChatGPT’s for the same period; other data says more than 30% of users revisited on Day 7 after first use. Guggenheim analyst Michael Morris, quoting Apptopia, says Muse averaged 1.3 million daily actives in five weeks—a mere 0.05% of Meta’s 3.6 billion daily actives. He expects Meta to prioritize expanding usage rather than short-term monetization.

However, monetization has only just started: as of October 12, Meta started charging for Muse, launching $20 and $100 monthly subscriptions with the new model—no revenue data yet. There are also reports that Muse is blocked from some websites, including Amazon—highlighting the engineering difficulty in building a truly reliable agent.

Free cash flow: The real controversy of this report

If advertising is the ace, and Muse is the option, free cash flow is the ugliest page of this financial report.

Q2 operating cash flow was $31.9 billion, but capex (including leases) hit $31.1 billion, so free cash flow was just $784 million—down from $8.549 billion a year earlier. Q3 could be worse; Meta’s annual guidance for capex is $130–145 billion, and $50.9 billion was already spent in the first half of the year, so H2 capex will be $79–94 billion. If operating cash flow stays at $32 billion per quarter, H2 free cash flow will be between negative $15–30 billion. Surveys of institutions are blunt: Q3 free cash flow is -$9.14 billion, capex $40.97 billion.

Where’s the money coming from? From borrowing. In Q2, Meta issued $24.9 billion in long-term debt, ending the quarter with $83.7 billion in LT obligations and $90.3 billion in cash and securities; there was no share buyback in Q2 (vs $10.2 billion a year ago). CFO Susan Li clearly stated on the Q2 call that Meta intends to use “a greater proportion of debt” for long-term infrastructure. Depreciation is also accelerating: Q2 depreciation was $6.4 billion, up 46% y/y, while capex was about five times that; with new data centers coming online, depreciation will further rise. Based on forward models (16% tax rate, about 2.57 billion diluted shares), every $10 billion increase in annual depreciation cuts EPS by about $3.3.

Competition

A new factor has emerged. Since October 8, Meta has stopped accepting ads from ByteDance on Facebook, Instagram, etc., affecting seven countries including the US, Canada, Egypt, Indonesia, Japan, Thailand, and Vietnam; the restriction also applies to third-party ads that link to TikTok and other ByteDance products. A Meta spokesperson was direct: “We are not obligated to accept advertising from a competitor whose purpose is to take our users away.” TikTok previously removed direct profile links to Instagram.

Valuation and Trading

According to StockAnalysis, based on the consensus 2027 EPS of $34.55, three scenarios: Bear case P/E 18x (market treats Meta as capital-intensive), Base 22x (about current price), Bull 26x (Muse and planned cloud business become real revenue sources). In addition, Meta Q2 free cash flow yield was just about 0.2%, and dividend yield about 0.3%—looks cheap on valuation, but there is little real cash return for shareholders now.

Wall Street’s disagreements center on price targets. The bearish-on-spending camp: Goldman Sachs cut its target from $860 to $762, citing potential upward pressure on 2027 opex—though it also raised its 6- to 18-month revenue forecast and kept the Buy; Cantor Fitzgerald held Overweight but trimmed its target from $860 to $840.

Bullish-on-monetization camp: UBS raised its target from $715 to $884, arguing for the launch of Business Agents, Muse usage growth, as well as WhatsApp and Messenger monetization and two generative AI monetization paths via search ad share gains; Guggenheim from $700 to $850; BNP Paribas' Nick Jones at $885; Pivotal Research from $790 to $930; both Truist and Wells Fargo at $1,000 (the latter also warns AI costs could cap 2027 profits). Bank of America analyst Justin Post maintains Buy with $810 target.

As for the stock trajectory, Meta rose from about $550 in August to around $780 in September, then entered a narrow $60 range. On technicals, resistance is around $780–786 (a breakout could open space above $800), support around $710–732. Recent path: October 9 close $718.67 → Oct 12 close $747.72 (+4.04%) → Oct 13 close about $763 → Oct 14 fell about 3% after Goldman’s target cut, etc. On Thursday, Meta closed down 0.11% at $736.72.

Options markets imply a ±4.1% move for this earnings event, while Meta’s average absolute earnings-day move over the last eight quarters was 7.4%—different metrics but pricing seems restrained. Another statistic to remember: over the past eight quarters, Meta shares rose after earnings just four times—a beat alone doesn’t guarantee gains. Pepperstone’s Chris Weston set a higher bar: revenue should meet or exceed $64 billion, margins should beat forecasts, Q4 guidance must be positive, and there must be concrete proof of AI monetization to reinforce the bull thesis; otherwise, soft guidance or yet another sharp capex hike could trigger a selloff even if Q3 is a beat.