AI Investment Cycle Starts to Pay Out — But Zuckerberg Just Raised the Ante on Capex

Results vs. Consensus

Quality of the Beat

• Revenue: Almost entirely organic and ad-driven. Family of Apps ad revenue was +16% reported, +20% constant currency — currency was a 3-point headwind that should ease into Q2 (guided to a 1% tailwind). Online commerce was the largest vertical contributor; Rest of World and North America led at +19% and +18% respectively.
• Margins: The 41.5% operating margin (+330bps YoY) was driven primarily by an unusually low 9% effective tax rate (excess tax benefits from share-based comp) and a 34% drop in G&A on lower legal accruals — both partially one-time. R&D was up 22% reflecting the ramping infra and AI talent investment, so the underlying operating leverage is real but flattered. Stripping out the tax-rate benefit, the EPS beat would be roughly $5.85 vs. consensus $5.23 — still a clean beat, just narrower.
• EPS: $6.43 of GAAP diluted EPS (+37% YoY) versus $5.23-5.28 consensus is a wide miss-of-the-miss for the bear case. Diluted share count fell to 2,590M (from 2,625M), so buybacks ($13.4B in the quarter) are doing real per-share work.
• FCF miss: The only blemish — operating cash flow of $24.0B was strong, but the $12.9B in property/equipment purchases plus $0.75B in finance lease principal payments compressed FCF to $10.3B vs. $12.5B last year. This is mechanical to the capex acceleration, not a working-capital problem.

Segment Performance

Family of Apps

FoA is doing exactly what it needs to do: monetizing the AI investment cycle through pricing, not just impressions. The +10% price-per-ad print versus +5% impression growth is the clean tell — pricing acceleration on the back of conversion-rate gains is what hyperscalers see when their ranking models genuinely improve. Susan Li flagged that "year-over-year conversion growth remains strong, and in fact, we continue to see conversions grow at a faster rate than ad impressions in Q1," which is the cleanest framing of why advertisers are paying up.

"We're focused on building full general intelligence. All of the opportunities that I've discussed today are downstream of delivering general intelligence and doing so efficiently." — Mark Zuckerberg, CEO

The "Other revenue" line (+34% to $510M) is a small number with disproportionate strategic importance — WhatsApp Business and Meta Verified are the early monetization vectors for messaging and the subscription framework Meta will need post-DMA in Europe. Threads at 350M MAUs and now globally monetized (ads opened to all advertisers in 30+ markets in April) is not yet material but is the next surface to watch.

Assessment: The cleanest quarter of ad-side execution since the 2022 efficiency cycle began. Pricing is doing the heavy lifting, which is sustainable as long as conversion lift continues.

Reality Labs

RL is the one segment where the investment cycle remains entirely a cost story. Revenue down 6% to $412M with a $4.2B operating loss (essentially flat YoY) — the bear case here is unchanged. The constructive read is that Ray-Ban Meta unit volumes are running 4x YoY and Zuckerberg explicitly framed the third-generation hardware target as 10M units, with new launches with EssilorLuxottica coming "later this year." The bear read is that quarterly losses have been remarkably consistent at ~$4B for four straight quarters with no inflection visible.

"By the time they get to their third generation, [leading consumer electronics products] are often selling 10 million units and scaling from there. I'm not sure if we're going to do exactly that, but I think that that's like the ballpark of the opportunity." — Mark Zuckerberg, CEO

Assessment: RL remains a "fund it from FoA cash flow and trust the long arc" story. The Ray-Ban Meta data is the first thing that has actually moved in 18 months; until unit volumes meaningfully scale, the segment is a tax on Family of Apps returns.

Key KPIs

The price/impression mix shift is the single most important number in the table. Through 2023-24, the story was "impressions up, price flat" as Reels was monetizing below feed. In Q1 2025 it has decisively flipped to a price-led model, which is where ad-platform value compounds.

Key Topics & Management Commentary

Overall Management Tone: Confident on AI execution, deliberately matter-of-fact on the capex raise, and notably more specific on the EU DMA risk than the prior call. Zuckerberg leaned into the "five opportunities" framing (improved advertising, more engaging experiences, business messaging, Meta AI, AI devices) which is now the running scorecard he expects to be measured against. Li was crisp on guidance and restrained on speculation — she explicitly declined to discuss 2026 capex.

The GEM Ad-Ranking Model

GEM (Generative Ads Recommendation Model) is the headline AI deliverable of the quarter and the one with the cleanest near-term P&L tie-out. Susan Li disclosed it is "twice as efficient at improving ad performance for a given amount of data and compute" and that initial Reels tests show up to a 5% lift in ad conversions. This is the third generation of model architecture changes (after Lattice and Andromeda) that Meta has rolled out in roughly 18 months — and the cumulative effect on the price-per-ad line is now visible.

"Year-over-year conversion growth remains strong, and in fact, we continue to see conversions grow at a faster rate than ad impressions in Q1, reflecting increased conversion rates." — Susan Li, CFO

Assessment: GEM is the most important technical disclosure on the call. If the Reels-test conversion lift generalizes across feed and Stories surfaces in 2H, the price-per-ad line could re-accelerate even if impression growth flattens. This is the bull case in mechanical form.

Capex Raise: $64-72B from $60-65B

The $4-7B raise to the 2025 capex range is the most contested number on the call. Li attributed it to two drivers: (1) accelerated data-center build-out as Meta flexes its build strategy to bring capacity online faster in 2025-26, and (2) higher infrastructure hardware costs reflecting tariff/sourcing uncertainty. Importantly, she declined to break out the split.

"Even with the capacity that we're bringing online in 2025, we are having a hard time meeting the demand that teams have for compute resources across the company." — Susan Li, CFO

The bear framing is that this is the fourth consecutive capex raise in 18 months and that "demand from internal teams" is not the same as ROI on external monetization. The bull framing is that ad-side AI investments (GEM, Andromeda) are demonstrably converting compute into pricing, and that the marginal dollar of capex is funded entirely by the $24B of operating cash flow Meta generated in the quarter.

Assessment: The capex raise is large enough to matter, but the cash-flow math still works comfortably. The legitimate bear point is forward-year ambiguity — Li explicitly refused to discuss 2026, which is the right disclosure posture but does not eliminate the question.

EU DMA Subscription-Model Decision

This is the most quantifiable near-term risk on the call. The European Commission has ruled Meta's subscription-for-no-ads model non-DMA-compliant. Meta will appeal but expects to need to make modifications that "could result in a materially worse user experience for European users and a significant impact to our European business and revenue as early as the third quarter of 2025." Susan Li disclosed that European Economic Area + Switzerland advertising revenue was 16% of 2024 worldwide total revenue.

"It is really too early to speak about what those changes could be because we are in the process of engaging with the European Commission." — Susan Li, CFO

Assessment: 16% of advertising revenue is a meaningful exposure. A "significant impact" is not quantified and likely will not be until the Q2 call. The Q2 guide range of $42.5-45.5B does not appear to bake in the DMA hit; investors should expect the Q3 outlook (delivered in late July) to be the one where this is actually priced.

Llama 4 and the Standalone Meta AI App

Llama 4 launched earlier in April with a 17B-parameters-per-expert architecture explicitly designed for low-latency voice and Meta's own infrastructure shape. Zuckerberg emphasized the strategic logic of in-house model development: control of destiny, optimization for Meta's compute footprint, and the ability to distill smaller production models from larger ones (Behemoth) — a capability not available with closed-source partner models.

Meta AI now has nearly 1B MAUs across the family of apps, with the standalone Meta AI app launched in Q1 in the U.S. and Canada. Zuckerberg framed monetization as 12+ months out: "I expect that we're going to be largely focused on scaling and deepening engagement for at least the next year, before we'll really be ready to start building out the business here."

"There will be a large opportunity to show product recommendations or ads, as well as a premium service for people who want to unlock more compute for additional functionality or intelligence." — Mark Zuckerberg, CEO

Assessment: Meta AI is currently a cost line; the call does not change that. The interesting disclosure is the explicit articulation of two future revenue streams (ads-in-AI and a premium subscription) — this is the most concrete monetization roadmap Zuckerberg has provided and signals 2026 is the year these become real.

WhatsApp Business and the AI-Agent Vision

Zuckerberg's most aggressive pitch on the call was for business messaging as the "next pillar of our business." The framing: every business will soon need an AI agent the way they now need a website, and Meta's messaging surfaces (WhatsApp 3B MAUs, Messenger 1B+ MAUs) are the natural distribution. He highlighted Thailand and Vietnam as countries where messaging-led commerce already drives top-10 revenue contribution despite ranking in the 30s by GDP, and argued AI economics now make this model exportable to developed markets.

Assessment: This is the most credible long-duration growth vector beyond Family of Apps ads. The numbers are not yet material, but the strategic logic is sound and the testing is broadening.

Guidance & Outlook

The Q2 revenue range of $42.5-45.5B is unusually wide ($3B), which Li attributed to "the dynamic macro environment" — specifically reduced U.S. ad spend from Asia-based e-commerce exporters anticipating the May 2 de minimis exemption expiration. The midpoint of $44B implies +14% YoY growth, decelerating from Q1's +16% — which is still respectable but reflects the macro caution.

The expense reduction is small ($1B at midpoint) and reflects refined headcount-comp forecasts partially offset by higher infra costs. The capex raise is the meaningful change.

Implied 2H ramp: If Q2 lands at midpoint ($44B), full-year revenue would need ~$176-180B for low-teens growth — implying H2 around $89-93B, or roughly +13-17% YoY. Achievable, with conviction-builders being GEM rollout and Threads ads ramp.

Street at: Pre-print consensus was around $43.8B for Q2; Meta guided to a midpoint of $44B, which is in line. The market will be watching whether the EU DMA hit shows up in the next guide.

Guidance style: Conservative on revenue, deliberately ambiguous on 2026 capex, surprisingly disciplined on FY expenses (a small tightening despite the AI ramp).

Analyst Q&A Highlights

AI & Meta AI Strategy

• Brian Nowak, Morgan Stanley: Asked about Llama strategic priorities and Meta AI U.S. traction. Zuckerberg leaned hard on the in-house AI thesis: low-latency voice, long context windows, and the ability to distill production models from larger ones. Li noted WhatsApp is the largest entry point for Meta AI engagement, with Facebook second.
  Assessment: Clear articulation of why Meta is building its own foundation models. The U.S. traction question was deflected — likely because U.S. share of Meta AI MAUs is still modest.
• Eric Sheridan, Goldman Sachs: Asked about the standalone Meta AI app rationale vs. the integrated experience. Zuckerberg said the standalone app is "particularly important in the United States" given iMessage's dominance over WhatsApp domestically.
  Assessment: Honest acknowledgment that Meta cannot rely on WhatsApp distribution for Meta AI in the U.S.
• Youssef Squali, Truist: Asked whether the AI assistant market is winner-take-most or fragmented. Zuckerberg argued for fragmentation by use case, with personalization as the key differentiator.
  Assessment: Reasonable framing but probably understates how concentrated leadership in personal AI may become.

Capex & Infrastructure

• Justin Post, Bank of America: Asked how the ROI works without enterprise demand to anchor the build. Li argued internal demand from ad-ranking and recommendation teams already exceeds available compute.
  Assessment: This is the right framing — Meta's capex is funding its own ads platform, not third-party customers. ROI is direct.
• Doug Anmuth, JPMorgan: Asked for a split between data-center spend and hardware-cost increases, plus the rumored AI-infra partnership conversations. Li declined to break out the split and said Meta is "funding the infrastructure that is being used to train Llama" itself.
  Assessment: Useful denial of any meaningful capex offload via partnerships.
• Kenneth Gawrelski, Wells Fargo: Asked whether the 2025 capex range implies anything about 2026. Li explicitly declined to comment.
  Notable dodge: This is the right disclosure posture but means the 2026 capex print, when it comes, is unbounded on the upside.

Macro & Vertical Mix

• Justin Post, BofA: Asked about Asia-based e-commerce supply softness. Li acknowledged "reduced spend in the U.S. from Asia-based e-commerce exporters" tied to the May 2 de minimis change, with some redirection to other markets.
  Assessment: Real but contained — the broader April trend was "generally healthy."
• Mark Mahaney, Evercore: Asked about other vertical softness and whether Reality Labs losses can come down. Li flagged gaming and government/politics as soft (China-game lap and post-election normalization). Zuckerberg deflected on RL losses, pointing to scaling unit volumes for Ray-Ban Meta.
  Assessment: Reality Labs answer is unsatisfying — 4x unit growth is real but on a tiny base; the operating loss line is unchanged.

Reality Labs Trajectory

• Mark Mahaney, Evercore: Asked when RL losses come down. Zuckerberg said the company is "definitely focused on doing the work more efficiently, but also very optimistic about what we're seeing," especially on Ray-Ban Meta. He framed the third-gen target as ~10M units.
  Notable non-answer: No timeline on when losses moderate. The math implies RL losses widen before they narrow if hardware unit volumes scale.

What They're NOT Saying

1. 2026 capex: Li explicitly declined to discuss. Given the trajectory of upward revisions in 2024-25 and Zuckerberg's "demand exceeds supply" framing, the implicit signal is that 2026 capex will be materially higher than 2025. The market will price this when guided, not before.
2. EU DMA quantification: Meta has flagged "significant impact" but refused to quantify. This is appropriate while negotiations continue but means Q3 is the quarter where this likely becomes a hit-the-print event.
3. Reality Labs path to profitability: Zuckerberg deflected on Mark Mahaney's direct question. There is still no articulated path or timeline.
4. U.S. Meta AI engagement metrics: Zuckerberg deflected Brian Nowak's question on U.S. Meta AI traction. The 1B MAU global number is real but skewed by WhatsApp's non-U.S. footprint.
5. Threads monetization timeline: Ads opened to all advertisers in 30+ markets in April but management explicitly said Threads is "not expected to be a meaningful driver of overall impression or revenue growth in 2025." Real runway is 2026.
6. Andromeda mention frequency: Andromeda was the marquee ad-ranking model touted in 2024. On this call, GEM took center stage and Andromeda received only a passing mention as part of "prior model architecture improvements." Either Andromeda's lift has saturated or GEM is meaningfully better.

Market Reaction

• After-hours move (April 30): Up roughly 5% on the print, on the back of the EPS beat and contained Q2 guide.
• Volume: Elevated, in line with prior earnings prints.
• Analyst tone: Generally constructive, with the capex raise the only real friction point.

The post-print reaction is rational: the ad-side beat and Q2 guide are clean, the capex raise is digestible at current cash-flow generation, and the EU DMA hit is acknowledged but not yet quantified. Stocks tend to digest "good quarter, raised investment, untaxed risk" prints in the +3 to +6% range — Q1 is squarely in that zone.

Street Perspective

Debate: Is the Capex Raise Justified or a Black Hole?

Bull view: The marginal capex dollar is clearly producing measurable ad-side conversion lift via GEM, Andromeda, and recommendation-system improvements. Operating cash flow ($24B in Q1 alone) easily funds the $64-72B annualized capex range, leaving room for buybacks ($13.4B in Q1) and dividends ($1.3B). Until the conversion-rate gains stop showing up in price-per-ad, the spend is paying for itself.

Bear view: This is the fourth consecutive capex raise in 18 months. Internal-demand framing means Meta is essentially expanding without an external monetization anchor (unlike the hyperscalers, which can point to enterprise GenAI demand). 2026 capex is unbounded and Susan Li refused to talk about it. Llama monetization is 12+ months out at minimum.

Our take: The bull case has cleaner near-term evidence (the +10% price-per-ad line, the GEM 5% conversion lift). The bear case is a forward-year question that gets answered at the Q4 print when 2026 guidance lands. We are willing to trust the run rate through 2025.

Debate: Reality Labs — Tax or Optionality?

Bull view: Ray-Ban Meta unit volumes 4x YoY, the EssilorLuxottica refresh is coming, and Zuckerberg has now publicly anchored to a 10M-unit third-gen target. AI glasses are the strongest demand signal Meta has had in hardware in five years.

Bear view: RL has lost ~$4B every quarter for four straight quarters with no visible inflection. Quest sales are declining, which removes the volume base the segment was supposed to be built on. The Ray-Ban data is on a tiny base.

Our take: RL is currently a tax. The Ray-Ban inflection is real and worth watching, but the next 12 months are still loss-deepening. We will not give Meta credit for RL until the operating-loss line bends.

Debate: How Big a Hit Is the EU DMA Decision?

Bull view: 16% of 2024 advertising revenue is the maximum exposure, and the realistic hit is a fraction of that. Meta has navigated GDPR, ATT, and prior DMA pressure without structural revenue impairment.

Bear view: Susan Li explicitly used the word "significant" — that is unusual disclosure language for Meta and implies the impact will be felt. With timing flagged as "as early as the third quarter," this becomes a 2H 2025 print risk.

Our take: The bear has the better near-term case here. We model a ~150-300bps revenue-growth headwind starting Q3, embedded but not quantified.

Model Update Needed

Valuation impact: Slight upward revision to NTM revenue and EPS, partially offset by higher capex and a discrete EU adjustment. We lean modestly positive on intrinsic value.

Thesis Scorecard Post-Earnings

Overall: Thesis strengthened on the ad side, unchanged on RL, and bear point #3 (regulatory) was upgraded from latent to explicit risk. Net: bull case has more weight than bear case at current valuation.

Action: Initiating coverage at Outperform. Monitor (1) GEM conversion lift in Q2 print, (2) EU DMA quantification on Q2 call, (3) Q4 print for 2026 capex guide.