Superintelligence Pivot Lands With a 22% Revenue Print to Back It Up — Now the Capex Bill Comes Due

Results vs. Consensus

Quality of the Beat

• Revenue: The cleanest beat of the quarter. Constant-currency growth equals reported growth (FX neutral), so the +22% is fully organic. Family of Apps ad revenue of $46.6B was +21% (or +22% cc), with growth re-accelerating across geographies — Europe led at +24% (notable given the EU regulatory pressure), Rest of World +23%, North America +21%, APAC +18%. Online commerce was the largest contributor.
• Margins: 43% operating margin is genuinely impressive given R&D was +23% YoY. The drivers: cost of revenue +16%, marketing +9% (with employee-comp offset), G&A -27% on continued legal-cost reversals, and revenue growth of +22% providing operating leverage. Effective tax rate 11% was flat YoY (no repeat of the Q1 share-based-comp benefit). Q2 EPS quality is materially better than Q1 — the leverage was operational, not tax-rate aided.
• EPS: $7.14 vs. ~$5.88 consensus is a 21% beat with very high quality. Diluted share count fell to 2,570M (from 2,610M last year), continuing the buyback-driven per-share tailwind.
• FCF miss: The headline FCF of $8.55B missed by ~$2.5B versus the buyside whisper, but the miss is mechanical — capex stepped up to $17B in the quarter (versus $13.7B in Q1), and Meta committed $15.1B to non-marketable equity investments (Scale AI). Operating cash flow of $25.6B was strong; the cash conversion compression is intentional, not deteriorating.

Segment Performance

Family of Apps

FoA delivered the strongest quarter of organic growth since 2021. Importantly, the volume/price mix shifted from Q1's price-led model: impression growth re-accelerated to +11% (from +5% in Q1) on engagement tailwinds plus modest ad-load optimization on Facebook, while price-per-ad held at +9% (versus +10% in Q1). The combination — both lines now contributing meaningfully — is what produces +22% ad revenue. Susan Li's framing was that GEM (the new Generative Ads Recommendation system) is now the ranking-stage workhorse, with Q2 enhancements driving "approximately 5% on Instagram and 3% on Facebook Feed and Reels" conversion lifts. Andromeda (the retrieval-stage model) drove "nearly 4%" higher conversions on Facebook Mobile Feed and Reels. Lattice (the broader ranking architecture) drove another "nearly 4%" lift. Layered, these are not additive but the cumulative effect is visibly in the price line.

"Over the last few months, we've begun to see glimpses of our AI systems improving themselves. And the improvement is slow for now, but undeniable, and developing superintelligence... is now in sight." — Mark Zuckerberg, CEO

Family of Apps "Other" revenue jumped +50% to $583M, accelerating from Q1's +34%. Susan Li attributed this to WhatsApp paid messaging revenue growth and Meta Verified subscriptions. Click-to-message U.S. revenue was up 40%+ YoY — small absolute dollars but validating the long-arc messaging-as-pillar thesis.

Assessment: The cleanest segment quarter Meta has produced in three years. The volume/price duality is what hyperscalers see when their ranking infrastructure is genuinely compounding.

Reality Labs

RL revenue of $370M was actually +5% YoY (versus -6% in Q1), helped by accelerating Ray-Ban Meta demand "still outstripping supply for the most popular SKUs despite increases to our production earlier this year." Quest sales remained the offset. Operating loss widened modestly to $4.5B (vs. $4.2B in Q1). Zuckerberg announced the Oakley Meta HSTN — a new performance/sport AI glasses SKU launching Q3 — and committed to material disclosure at the Connect event in September.

"AI glasses are going to be the main way that we integrate superintelligence into our day-to-day lives." — Mark Zuckerberg, CEO

Assessment: The Ray-Ban supply-constrained framing is the most positive RL signal in 18 months. Connect on September 17 will be the next datable catalyst. RL operating losses are not coming down in the near term, but the unit economics on AI glasses (if they scale to even 5-10M units in 2026) start to look meaningfully different from the Quest run rate.

Key KPIs

The KPI line that matters most is the impression re-acceleration to +11% combined with sustained +9% pricing. That is the rare "engine firing on both cylinders" quarter. Instagram video time +20% YoY (with similar U.S. dynamics on Facebook) underpins the engagement tailwinds.

Key Topics & Management Commentary

Overall Management Tone: Materially more confident than Q1 — Zuckerberg used "superintelligence" 11 times and made it the explicit framing for the entire investment cycle. Susan Li was crisper on capex/opex, willing to give an early read on 2026 (which is unusual for her), and explicitly acknowledged the "massive amount of capital" being committed. The tonal shift from "we're scaling" (Q1) to "we are building the world's first gigawatt-plus cluster" (Q2) is the defining narrative pivot of the quarter.

The Superintelligence Pivot & Meta Superintelligence Labs

The most consequential strategic announcement of the call was the formalization of Meta Superintelligence Labs. The leadership: Alexandr Wang (formerly Scale AI) leading overall; Nat Friedman (formerly GitHub CEO) leading AI Products and Applied Research; Shengjia Zhao (ex-OpenAI) as Chief Scientist. Meta's $14.3B investment in Scale AI underwrites the data-supply backbone. Zuckerberg's framing — "we believe that personal superintelligence is now in sight" — is the most aggressive long-duration AI thesis any megacap CEO has stated.

"Meta has all of the ingredients that are required to build leading models and deliver them to billions of people. The people who are joining us are going to have access to unparalleled compute as we build out several multi-gigawatt clusters. Our Prometheus cluster is coming online next year, and we think it's going to be the world's first gigawatt-plus cluster." — Mark Zuckerberg, CEO

Assessment: This is the most ambitious strategic articulation Zuckerberg has made since the Metaverse pivot — but with a critical difference: this one is being funded by $200B+ in run-rate revenue from a healthy ad business, whereas the Metaverse pivot was funded against a deteriorating ad backdrop. The risk profile is asymmetric in a way the 2022 pivot was not.

Capex: Now $66-72B for 2025, With 2026 Set for Another Step-Up of Similar Magnitude

Susan Li tightened the 2025 capex range to $66-72B (from $64-72B) and explicitly previewed that 2026 capex will see "another year of similarly significant capital expenditures dollar growth." With 2025 midpoint at $69B and 2024 capex of ~$39B, the YoY dollar increase was ~$30B; another similar-dollar step-up implies 2026 capex in the $96-100B+ range. This is the headline number that will define Meta's 2026 narrative.

"We currently expect another year of similarly significant capital expenditures dollar growth in 2026 as we continue aggressively pursuing opportunities to bring additional capacity online to meet the needs of our artificial intelligence efforts and business operations." — Susan Li, CFO

Assessment: This is the cleanest forward-capex disclosure Meta has provided. The bull case is that the ad-side ROI on the prior year's capex is demonstrably working (GEM, Andromeda, Lattice all driving conversions). The bear case is that 2026 depreciation alone — flagged as the "single largest contributor" to expense growth — will compress operating margins back toward the 35-37% range.

EU LPA: Risk Got Bigger and Sooner

The Q1 framing was "modifications to our subscription-for-no-ads model could result in a significant impact... as early as Q3." The Q2 framing is sharper: the European Commission is now negotiating against the Less Personalized Ads (LPA) offering Meta introduced in November 2024, and Susan Li flagged that further EC modifications "could have a significant negative impact on our European revenue, as early as later this quarter" — meaning Q3, which is currently underway. Europe was 24% of ad revenue growth in Q2 (Meta's strongest geo).

"As the EC provides further feedback on LPA, we cannot rule out that it may seek to impose further modifications to it that would result in a materially worse user and advertiser experience." — Susan Li, CFO

Assessment: The escalating language matters. "Cannot rule out" combined with "as early as later this quarter" is a more direct disclosure than Q1. Investors should price a meaningful European-revenue headwind starting in Q3, even if not yet quantifiable.

Open Source: Quietly Hedged

Doug Anmuth pressed Zuckerberg on whether Meta's open-source posture has changed given the superintelligence framing. Zuckerberg's answer was unusually qualified: yes, Meta will continue to open-source some models; but as models get larger and more capable, "we would kind of wrestle with whether it's productive or helpful to share that or if that's really just primarily helping competitors." He also flagged "a whole different set of safety concerns" as Meta approaches superintelligence.

Assessment: This is a meaningful shift in tone. The Llama 1-3 open-source posture was an unambiguous "release everything." The superintelligence-era posture is "selective open source." This complicates the bull thesis around Llama as a developer-community moat, though it is the right disclosure given competitive realities.

Threads & WhatsApp Ads Surface Expansion

Threads ads went global to most countries (including the U.S.) in May for video and image campaigns, and Meta began introducing ads in WhatsApp Updates tab and channel subscriptions — the first systematic monetization of WhatsApp's 1.5B+ daily Updates-tab users. Susan Li was explicit that neither will be a "meaningful contributor to total impressions or revenue growth for the next few years," and that WhatsApp Updates ads will earn lower CPMs than Facebook/Instagram for the foreseeable future.

Assessment: 2025-26 are surface-build years; meaningful revenue contribution is 2027+. The strategic logic is sound (vast underutilized inventory) but should not be in near-term models.

Guidance & Outlook

The Q3 guide of $47.5-50.5B (midpoint $49B) implies +20% YoY growth — modestly decelerating from Q2's +22% but well ahead of consensus near $46B. The Q4 directional commentary (slower than Q3) reflects lapping a stronger Q4 2024, not a fundamental deceleration.

Implied Q3 ramp: $49B midpoint suggests ad-driven seasonality plus continued GEM/Andromeda lift. Achievable absent the EU LPA hit, which is the main downside risk.

Street at: Pre-print Q3 consensus was ~$46B; Meta guided ~$49B midpoint, so the print produces a ~6% guide-up beat.

Guidance style: Conservative on near-term revenue (Meta typically beats the high end), aggressive on 2026 capex/opex disclosures. The 2026 preview is unusually early and unusually direct.

Analyst Q&A Highlights

Superintelligence Strategy & Talent

• Eric Sheridan, Goldman Sachs: Asked what learnings from the past 3-6 months have driven the talent-and-compute strategy shift. Zuckerberg argued the "more aggressive assumptions on AI timelines have been the ones that most accurately predict what would happen" and emphasized his conviction that small, talent-dense teams beat large research orgs for frontier work.
  Assessment: Honest framing. The "small team" structure is a deliberate departure from Meta's typical large-team approach and signals he is willing to pay a premium for elite talent density.
• Brian Nowak, Morgan Stanley: Asked about the technological gating factors Meta is most focused on. Zuckerberg pointed to AI self-improvement as the central research priority — "you're not just going to be learning from people because you're trying to build something that is fundamentally smarter than people."
  Assessment: Most ambitious technical articulation Zuckerberg has made on a call. Worth taking seriously even if it remains aspirational.

Capex & Financing

• Doug Anmuth, JPMorgan: Pressed on whether 2026 capex implies more than $100B and whether external financing partners are in play. Susan Li acknowledged Meta is "exploring ways to work with financial partners to co-develop data centers" — first time Meta has confirmed that conversation publicly.
  Assessment: Material disclosure. Co-development structures could meaningfully reduce on-balance-sheet capex while preserving compute access.
• Justin Post, Bank of America: Asked whether the capacity could ever be externalized. Susan Li said internal needs absorb everything currently in the build pipeline; external monetization is not on the table.
  Assessment: Reaffirms that Meta's capex is funding its own ad and AI products, not building a third-party hyperscaler business.

Open Source AI

• Doug Anmuth, JPMorgan: Asked whether Meta's open-source AI posture has shifted. Zuckerberg's answer was notably qualified — Meta will continue open-sourcing "some" models, but the largest frontier models may be too capable or too sensitive to release.
  Notable shift: First public hedge on the open-source commitment Meta has made loudly for two years.

Core Recommendation Engine

• Brian Nowak, Morgan Stanley: Asked what's still ahead for engagement gains. Susan Li flagged session-adaptive recommendations, smaller-creator content surfacing, niche-interest exploration, and cross-surface foundation models in the long-term roadmap.
  Assessment: Detailed product roadmap with multiple shots on goal — not relying on a single architecture upgrade.

What They're NOT Saying

1. Quantified 2026 capex/opex: Susan Li gave directional commentary but no specific 2026 dollar figures. The market will wait for the Q4 2025 print (Jan/Feb 2026) to lock these in. Until then, the implicit ~$100B 2026 capex range is doing the analytical work.
2. EU LPA quantification: Three months after Q1's first warning, Meta still has not put a dollar number on the European-revenue impact. The escalation in language ("cannot rule out... materially worse... as early as later this quarter") suggests the impact is becoming concrete; the absence of numbers means it is still being negotiated.
3. Reality Labs unit volumes: Zuckerberg referenced "demand outstripping supply" for Ray-Ban Meta but again declined to disclose unit numbers. This is the third quarter in a row where an interesting Ray-Ban data point is offered without a base. Connect on September 17 is the deadline for actual unit disclosure.
4. Scale AI synergy detail: The $14.3B Scale AI investment was disclosed but not framed in P&L terms — what does Meta gain operationally that justifies the cap-table commitment? The Wang/Friedman/Zhao leadership announcement implies it is more about talent and data access than equity returns.
5. Llama 4.1/4.2 release timing: Zuckerberg said "good progress" but no dates. With OpenAI/Google/Anthropic shipping aggressively, the Llama cadence question is becoming a competitive question.
6. U.S. tax law impact: Meta acknowledged the new U.S. tax law will reduce federal cash tax but said "we cannot quantify the magnitude at this time." This is a material EPS-and-FCF tailwind that is in the model but unsized.

Market Reaction

• After-hours move (July 30): +11-12% on the print, the strongest reaction Meta has had to an earnings beat in two years.
• Volume: ~2.5x average on the post-print session.
• Market cap impact: ~$200B added on the print, taking Meta's market cap above $1.97T.
• Analyst tone: Broadly constructive on the operational beat; capex framing is the friction point but most desks are willing to accept it given the demonstrable ad-side ROI.

The reaction is rational — a 6% revenue beat with 500bps of margin expansion is exactly what the bull case needs to validate Q1's initiation. The capex 2026 preview is a known overhang but the +12% move shows the market is willing to underwrite the spend if the ad business keeps compounding.

Street Perspective

Debate: Is the 2026 Capex Step-Up Investable?

Bull view: Q1 and Q2 each delivered standalone evidence that AI capex translates into measurable ad-side returns (GEM +5% lift, Andromeda +4% lift, Lattice +4% lift, all on top of normal engagement growth). Operating cash flow of $25.6B in Q2 alone funds the build. The Scale AI investment gives Meta a structural data advantage. Co-development financing is now publicly on the table.

Bear view: Another $30B+ step-up in 2026 capex pushes total spend to ~$100B with depreciation ramping sharply. Operating margin compression is mathematically nearly certain if revenue growth decelerates from Q2's +22%. This is a leveraged bet on superintelligence timelines — Zuckerberg's "now in sight" framing is the kind of disclosure that ages either very well or very poorly.

Our take: Q2 is the strongest single data point for the bull case to date. The question is whether the GEM/Andromeda/Lattice returns curve flattens before the depreciation step-up hits in 2026. We model margin compression to ~38% in 2026 but with absolute earnings growing on the revenue base.

Debate: Reality Labs Inflection?

Bull view: Ray-Ban Meta supply-constrained, Oakley Meta HSTN launching, Connect event on September 17 likely to bring multi-product disclosure including Hyperion (Meta's high-end AR demo). The AI glasses category is the first hardware product Meta has had genuine consumer pull for.

Bear view: Operating loss widened to $4.5B (from $4.2B). Quest sales declining. Without unit disclosure, "demand outstrips supply" is unfalsifiable.

Our take: Connect is the catalyst. Until then, RL remains a tax to be funded out of FoA cash flow.

Debate: How to Value the Superintelligence Optionality?

Bull view: If Meta's AI investments deliver even partial superintelligence in 3-5 years, the option value is enormous — autonomous agents replacing white-collar work, personalized AI as a recurring-revenue subscription, AI-generated content unlocking new ad inventory, etc. None of this is in current models.

Bear view: The Metaverse pivot showed the cost of overweighting Zuckerberg's long-duration tech bets. Superintelligence may simply be the new Metaverse, with $30B+/year of capex committed before any unit economics are validated.

Our take: The differences from Metaverse are real (active business funding it, demonstrable AI-side ad ROI), but bear-case discipline matters. We give zero credit for superintelligence-driven optionality in our model and underwrite Meta on the run-rate ad business plus probable monetization of Meta AI by 2027.

Model Update Needed

Valuation impact: Modest upward revision to NTM revenue and EPS, partially offset by 2026 margin compression. Net mildly positive on intrinsic value, with substantial 2026 modeling uncertainty.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened on the operational side; new bear point (2026 margin compression) added but is partially priced in by the capex 2026 preview itself. Net: bull case continues to dominate at current valuation.

Action: Maintaining Outperform. Connect on September 17 (RL update) and the Q3 print (EU LPA quantification, 2026 capex sharper) are the next catalysts.