META · Q2 2025 Earnings

Bullish

Headline numbers

EPS

Q2 FY2025

$7.14

Revenue

Q2 FY2025

$47.52B

+22.0% YoY

Gross margin

Q2 FY2025

82.1%

Free cash flow

Q2 FY2025

$8.55B

Operating margin

Q2 FY2025

43.0%

Key financials

Q2 FY2025

Metric	Q2 FY2025	YoY
Revenue	$47.52B	+22.0%
EPS	$7.14	—
Gross margin	82.1%	—
Operating margin	43.0%	—
Free cash flow	$8.55B	—

Guidance

Prior quarter data unavailable — comparison not possible.

Segment performance

Q2 FY2025

Segment	Q2 FY2025	YoY
Family of Apps	$47.146B	+21.7%
Reality Labs	$0.37B	+4.8%
Advertising Revenue	$46.563 billion	—
Advertising Revenue Growth YoY	21%	—

Platform metrics

Q2 FY2025

Segment	Q2 FY2025
Family Daily Active People (DAP)	3.48 billion
Ad Impressions Growth YoY	11%
Average Price Per Ad Growth YoY	9%

Management tone

Transcript prepared remarks were not available for this brief; tone analysis below is drawn entirely from the Q&A exchanges.

Management's posture in Q&A was unusually direct about the scale and duration of the AI capex commitment. Susan Li's framing of 2026 — "similarly significant capital expenditures dollar growth" and "expense growth rate expected above 2025" — is not a hedge or a "we'll update you" placeholder; it's a pre-commitment delivered before the fiscal year even begins. That confidence reads as either conviction about the ad business's ability to fund it, or an attempt to anchor investors before sticker shock.

Zuckerberg's repeated emphasis that "more aggressive AI timeline assumptions have been most accurate" is the tell. He is telegraphing that internal models point to faster superintelligence progress than consensus expects, and that the company is staffing and spending accordingly — including a deliberate shift to "small, elite talent-dense teams" for frontier research. This is a meaningful organizational change being articulated to investors, not just an HR detail.

On ROI, Li was explicit and unhedged: Gen AI will not drive "meaningful revenue this year or next." Paired with capex potentially exceeding $100B annually (a figure floated by JP Morgan and not pushed back on), this is the most honest framing of the investment/return gap any hyperscaler has offered this cycle. Management is asking investors to underwrite a 3+ year duration trade.

The exploration of "co-development partnerships for data centers" disclosed to JP Morgan is new and material. It signals that even Meta — with $8.5B quarterly FCF and a fortress balance sheet — is examining off-balance-sheet structures to fund the build. That's worth watching.

Q&A highlights

Eric Sheridan · Goldman Sachs

Asked about key learnings from AI strategy evolution over 3-6 months, shifts in talent acquisition and compute investments, and how superintelligence strategy has changed. Also asked about drivers of talent and compute investment impact on CapEx and optics over 12-18 months.

Mark emphasized that faster/more aggressive AI timelines have proven most accurate; cited internal examples like teams using Llama4 to build autonomous agents improving Facebook algorithm. Discussed importance of small, elite talent-dense teams for frontier research vs. larger teams for product optimization. Susan detailed 2026 expense drivers: infrastructure (depreciation acceleration, operating costs, cloud services) and employee compensation from AI hiring; outlined similar significant CapEx growth expected in 2026 driven by gen AI capacity scaling.

More aggressive AI timeline assumptions have been most accurate predictorsAutonomous AI agents already being used internally to improve engagement algorithms2026 infrastructure costs driven by sharp acceleration in depreciation expense growth2026 expense growth rate expected above 2025 rate

Doug Amuth · JP Morgan

Asked about Meta's stance on open-source AI as it pursues superintelligence and ROI on infrastructure investments. Also asked about financing strategy for CapEx potentially exceeding $100B annually.

Mark stated that Meta's open-source thinking hasn't changed; will continue open-sourcing some models but not everything, citing concerns about very large models not being practical for others to use and safety concerns approaching superintelligence. Susan confirmed Meta expects to finance majority of CapEx internally but is exploring partnerships with financial partners to co-develop data centers; no finalized transactions announced.

Will continue open-sourcing leading models but selective about very large modelsNew safety concerns around superintelligence inform open-source decisionsExploring co-development partnerships for data centers to provide flexibilityNo finalized transactions on external financing for infrastructure announced

Brian Nowak · Morgan Stanley

Asked about technological constraints and gating factors for superintelligence development over next 24 months compared to 12 months ago. Also asked about core engagement platform improvements expected in next 18 months.

Mark identified self-improvement and scaling paradigms as key research focus; emphasized importance of AI learning to improve itself rather than just learning from humans. Highlighted shift toward smaller, talent-dense teams for frontier research vs. larger teams for product optimization. Susan outlined near-term engagement improvements: more adaptive session-based recommendations, helping smaller creators break through, interest exploration. Long-term bets include foundational recommendation models across services, deeper LLM integration, and efficiency optimization.

Self-improvement is fundamental research area for superintelligence developmentShift toward smaller, elite talent-dense teams optimal for frontier researchNear-term: adaptive session recommendations, creator discovery, interest explorationLong-term: cross-service foundation models, deeper LLM integration in recommendations

Justin Post · Bank of America

Asked whether Meta's massive infrastructure spend is purely for internal use or could monetize external capacity. Also asked about ROI measurement on CapEx and long-term return expectations.

Susan stated current focus is entirely on internal capacity needs for core AI, ads systems, and frontier model training; no external use cases being considered currently. On ROI, confirmed strong measurement and returns on core AI side; acknowledged Gen AI side much earlier on return curve with no meaningful revenue impact expected this year or next; remain optimistic about medium-long term monetization through five pillars. Highlighted infrastructure designed with fungibility in mind for flexibility.

No current plans to monetize external capacity from infrastructure investmentsCore AI delivering strong, measurable ROIGen AI not expected to drive meaningful revenue in 2025-2026Five pillars identified as future monetization opportunities

Mark Shmulek · Bernstein

Asked about KPIs and markers Meta is tracking for superintelligence progress. Also asked whether relationship between revenue/core business performance and investment cadence has changed.

Mark identified internal KPIs: quality of teams/people, quality of models produced, rate of improvement of other AI systems, contribution of foundation models to all company systems. Noted standard playbook of translating technology to products that scale to billions before monetization, with built-in lag. Susan confirmed primary profitability focus is consolidated operating profit growth over time; acknowledged it won't be linear and that years with big investments will see lower profit growth; framed current moment as time for major AI investments to enable compelling future profit growth.

Internal KPIs: team quality, model quality, improvement rate of downstream AI systemsStandard playbook: research → product scaling → monetization with multi-year lagPrimary focus remains consolidated operating profit growth over timeProfit growth will be non-linear; will be lower in years of major investments

What to watch into next quarter

Q3 revenue landing within or above the $47.5–50.5B guide range — the low end implies flat QoQ, the high end +6% QoQ; anything below midpoint with FX as a stated tailwind would suggest underlying ad demand softening.

Ad pricing growth sustaining above high-single digits — Q2 was +9%; a drop to low-single digits would shift the growth mix to a less healthy volume-only profile.

FY2026 capex range disclosure — management has telegraphed "similarly significant" dollar growth; the actual number, when given, will determine whether 2026 capex lands closer to $80B or $100B+. This is the single most market-moving disclosure pending.

Reality Labs operating loss trajectory — revenue at $370M is immaterial; the loss line (not in this extraction) determines whether RL remains a tolerable drag or escalates.

Any concrete announcement on data-center co-development partnerships — Li opened this door in Q&A; first named partner or structure would be a meaningful capital-structure signal.

Q4 revenue deceleration magnitude — management flagged Q4 YoY growth will be slower than Q3 due to tougher comp; investor reaction depends on how slow.

Sources

Meta Q2 2025 Press Release (Form 8-K Exhibit 99.1), filed July 30, 2025 — https://www.sec.gov/Archives/edgar/data/1326801/000162828025036719/meta-06302025xexhibit991.htm
Meta Q2 2025 Earnings Call Q&A (transcript excerpts)

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.