Meta Platforms, Inc. (META) | The Buildout — AI Infrastructure
The Verdict
Meta Platforms sells advertising across its family of apps — Facebook, Instagram, WhatsApp, Messenger, Threads, and Meta AI — and is using that cash flow to become a full-stack AI company. It builds its own models, runs its own data centers, and buys large amounts of compute, power, and networking. In the AI buildout, Meta sits on the demand side: a hyperscaler-scale purchaser of AI hardware and capacity, and a developer of consumer and business AI products built on top of it. Management points to three areas — AI improving the core apps, personal agents, and a large enterprise push that could include selling compute.
| Market Cap | — |
| Revenue (TTM) | $228.2B |
| Revenue Growth | +27.7% |
| EBITDA Margin (TTM) | 48.0% |
| Net Debt | $22.1B |
| Earnings Beats | 7 of 7 |
| P/E (TTM) | — |
| EV/EBITDA (TTM) | — |
What We Like
- The advertising engine keeps compounding on a huge base: Q2 2026 total revenue rose 28% YoY, with Family of Apps ad revenue of $59.4B, up 27%.
- AI inside the core is producing measurable lifts, not just narrative: Advantage+ reached an over $75B annual revenue run rate, and user-understanding models combined with GEM and sequence learning drove an 8.3% increase in Facebook ad clicks and a 15.7% uplift in conversions.
- The engagement base is durable: 3.6B people use at least one Meta app daily, Instagram has 2B daily actives, and Threads crossed 500M monthly actives.
- New AI lines are scaling from a small base: business agents are globally available on WhatsApp and Messenger with more than 1M businesses using them weekly, and Meta AI daily interactions rose 60% after the Muse Spark rebuild.
- Compute and supply are being secured at scale: Broadcom is to deploy 3 GW of MTIA custom silicon through end-2028, AMD up to 6 GW of Instinct GPUs, and Meta signed cloud capacity with CoreWeave (about $21B through 2032) and Nebius ($27B, of which $12B is committed).
What We’re Watching
- Ad growth is decelerating: Family of Apps ad revenue slowed to +27% YoY in Q2 from +33% in Q1, and ad impressions growth eased to +14% from +19%, while price per ad held at +12% in both quarters.
- Free cash flow and capital intensity: Q2 capex hit $31.1B and free cash flow compressed sharply; FY2026 capex is guided to $130B–$145B, with no 2027 outlook yet.
- The legal and regulatory load is now in the P&L: Q2 carried $2.4B of legal charges, and management said U.S. youth-related trials this year may ultimately result in a material loss.
- New AI products are early and undisclosed: business agents were free for most businesses as of Q1, personal agents have not shipped, and none of the new lines is reported separately.
The thesis is intact but is now being tested by a new financial regime rather than by a product setback. The core ads franchise still grew 28% in Q2 and the AI tools inside it show real lifts, which supports the case that AI is improving the existing business. At the same time, ad growth is decelerating and the capital-intensity step-change — a $130B–$145B capex year, compressed free cash flow, and a $2.4B legal charge — means the return on the build is not yet measurable. The open question is whether the new AI products and the capacity being financed turn into disclosed revenue before depreciation and legal costs weigh on earnings.
Earnings Beat
Meta reported Q2 2026 revenue of $60.8B, up 28% year over year, on gross margin of 81.4%. GAAP operating income fell 8% YoY to $18.8B — a 31% operating margin — held down by $2.4B of legal charges and $1.2B of severance; excluding those items, operating income would have risen 9%. Net income was $15.8B and capital expenditures reached $31.1B.
| Metric | Q2 FY2026 | Q1 FY2026 | Q2 FY2025 | YoY |
|---|---|---|---|---|
| Revenue | $60.8B | $56.3B | $47.5B | +28.0% |
| Gross margin | 81.4% | 81.9% | 82.1% | -70bps |
| EBITDA | $25.1B | $28.9B | $24.8B | +1.4% |
| EPS | $6.18 | $10.44 | $7.14 | −13.4% |
| Family of Apps other revenue | $1B | $885M | n/a | +73% YoY |
| Advantage+ annual revenue run rate | over $75B | n/a | n/a | — |
we're getting a lot of offers for compute at a significant premium over what we paid for it.— Mark Zuckerberg, 2026-07-29
Management tone: Management's tone shifted toward AI infrastructure and new-product monetization while staying candid on near-term cost. Mark Zuckerberg opened with three AI areas and several product announcements, and Susan Li named the demand constraint — including in the core business — and disclosed the free-cash-flow compression without softening it. Asked to rank which AI area scales first, or to give a 2027 capex outlook, management declined and emphasized optionality and dynamic planning.
Management Guidance
As stated, management guided Q3 2026 revenue of $61B to $64B, assuming foreign currency is about a 1% headwind to year-over-year total revenue growth. FY2026 total expenses were guided to $165B to $169B, with the lower end raised to absorb the $2.4B legal charge recognized in Q2. FY2026 capital expenditures were guided to $130B to $145B including principal payments on finance leases, narrowed from a prior $125B to $145B. The remaining-2026 tax rate was raised to 15% to 17% from 13% to 16%, and FY2026 operating income was reaffirmed as above 2025. No specific 2027 capex outlook was given.
Trajectory
Revenue is growing fast on a very large base. Q2 2026 total revenue was up 28% YoY and up 8% from Q1's $56.3B. Underneath, the core is decelerating: Family of Apps ad revenue slowed to +27% YoY from +33% in Q1, as ad impressions growth eased to +14% from +19%; price per ad held at +12% in both quarters. Margins compressed on a GAAP basis — operating margin fell to 31% from 41%, and EBITDA margin fell to 41.3% from 51.3% — driven by legal charges, $1.2B of severance, higher infrastructure and AI-talent costs, and rising depreciation as capex stepped up. Gross margin held near 81–82%.
The Model
The model projects FY+1 revenue of $255.0B and EBITDA of $117.8B, a 46.2% EBITDA margin. For FY+2 it projects revenue of $299.0B and EBITDA of $139.9B, a 46.8% margin. The near-term figure is anchored by the advertising franchise and the AI tooling already scaled inside it, such as Advantage+. The FY+2 step assumes the new AI products — business agents, the model API, subscriptions, and any compute sales — and the capacity being built begin to contribute.
| Metric | FY2025 | Next FY (E) | Following FY (E) |
|---|---|---|---|
| Revenue | $201.0B | $255.0B | $299.0B |
| YoY Growth | — | +26.9% | +17.3% |
| EBITDA | $101.9B | $117.8B | $139.9B |
| EBITDA Margin | 50.7% | 46.2% | 46.8% |
Projections are the median of 5 independent model runs. The model’s revenue sits 0.3% above analyst consensus.
As stated, management guided Q3 2026 revenue of $61B to $64B, assuming foreign currency is about a 1% headwind to year-over-year total revenue growth. FY2026 total expenses were guided to $165B to $169B, with the lower end raised to absorb the $2.4B legal charge recognized in Q2. FY2026 capital expenditures were guided to $130B to $145B including principal payments on finance leases, narrowed from a prior $125B to $145B. The remaining-2026 tax rate was raised to 15% to 17% from 13% to 16%, and FY2026 operating income was reaffirmed as above 2025. No specific 2027 capex outlook was given.
What Could Go Right — and Wrong
- Core ad growth holds near Q2's 28% while price per ad keeps offsetting slower impression growth.
- A personal agent ships and scales, and business agents move from free to a paid model at volume.
- The standing compute offers convert into a signed compute-sale contract at a premium to cost.
- Meta begins disclosing AI product revenue — API, business agents, subscriptions, or AI glasses — as a separate line.
- FY2026 capex lands within the guided range and the built capacity turns into revenue-generating compute.
- Ad growth keeps decelerating while the committed cost and capex base continues to grow.
- Free cash flow stays compressed and the funding mix leans further on debt and partnerships.
- Legal outcomes — youth-related trials or teen-safety remedies — prove material or cut engagement.
- The new AI product lines stay undisclosed and immaterial longer than management implies.
- Memory or power inflation lifts 2027–2028 costs above the current guide.
Looking Ahead
Over the next twelve months the story turns on a handful of points. Meta guides Q3 2026 revenue and FY2026 capex, with no 2027 capex figure yet. The dated near-term event is Meta Connect on September 23, 2026, where more of the glasses lineup is expected. Management has pointed to a personal agent, broader model-API distribution, and possible compute sales, but has given no dates for any of them, and has kept AI product revenue undisclosed. The year also carries scheduled U.S. youth-related trials that management says may result in a material loss.
- September 23, 2026Meta Connect glasses lineup — More Meta Glasses and styles, first shipping with Muse Spark.
- Q3 2026Q3 2026 revenue report — Tests the $61B–$64B guide against slowing ad growth.
- End Q3 2026Headcount reduction flows through — Majority of the ~8,000 impacted expected off headcount.
- 2026Youth-related legal trials — Management says these may result in a material loss.
- 2H 2027Broadcom MTIA delivery starts — Initial 1 GW of custom silicon begins delivery.
- 2030–2031Entergy generation in service — North Louisiana combined-cycle units come online.
Financials
Annual Summary
| Metric | FY2024 | FY2025 | TTM | YoY |
|---|---|---|---|---|
| Revenue | $164.5B | $201.0B | $228.2B | +22.2% |
| Gross Margin | 81.7% | 82.0% | 81.8% | +35bps |
| EBITDA | $84.9B | $101.9B | $109.7B | +20.0% |
| EBITDA Margin | 51.6% | 50.7% | 48.0% | 90bps |
| Net Income | $62.4B | $60.5B | $68.1B | -3.1% |
| Free Cash Flow | $54.1B | $46.1B | $41.0B | — |
| Net Cash | — | — | — | — |
Key Ratios (Trailing)
- P/E TTM—
- EV/EBITDA TTM—
- EV/Revenue TTM—
- Price/FCF TTM—
- Gross Margin (TTM)81.8%
- EBITDA Margin (TTM)48.0%
- Net Margin (TTM)29.8%
- ROIC24.2%
- FCF Conversion37.4%
- SBC / Revenue11.0%
The Company
Meta Platforms reports two segments: Family of Apps — Facebook, Instagram, Messenger, Meta AI, Threads, and WhatsApp — and Reality Labs, which covers VR devices, AI glasses, and AR. Substantially all revenue comes from selling advertising placements on its family of apps to marketers, as the FY2025 10-K states. The company says AI is 'accelerating every major part of our core business,' and it is layering new products on top: the Muse models and public API, business agents, Meta AI, and AI glasses.
Meta owns 30 data center locations globally and leases others, with about 10 million square feet of owned and leased office and building space as of Dec. 31, 2025; its headquarters is in Menlo Park, California. The operating model is shifting toward capital intensity, and Meta is building a multi-vendor compute base of Broadcom custom silicon, AMD Instinct GPUs, NVIDIA systems, and Arm and Qualcomm CPUs, alongside long-dated cloud contracts and extensive power agreements.
Business Segments
Competitive Landscape
Meta competes on several fronts, as the source material describes. Its 10-K flags Apple, whose iOS changes limit how Meta and others in digital advertising can target and measure ads, and Google, which previously proposed phasing out third-party cookies in Chrome. It names TikTok as a competitive product that has reduced some users' engagement. Beyond advertising, the source material lists AI model competitors (OpenAI, Anthropic, Google Gemini/DeepMind, Microsoft), infrastructure competitors that also build cloud and custom silicon (Amazon, Microsoft, Google), and Amazon again in agentic commerce.
- Apple10-K says Apple's iOS changes limit Meta's and others' ability to target and measure ads.
- Google10-K says Google previously proposed phasing out third-party cookies in its Chrome browser.
- TikTok10-K names TikTok as a competitive product that has reduced some users' engagement.
- OpenAINamed in the source material among AI model competitors; not discussed in detail.
- AmazonNamed among infrastructure and agentic-commerce competitors; its Alexa for Shopping reaches 350M customers.
Supply Chain
Meta sits on the demand side of the AI supply chain — a hyperscaler-scale buyer of silicon, cloud capacity, power, and networking. Several counterparties name Meta directly in their own filings and calls, including Broadcom, AMD, CoreWeave, Nebius, Corning, Arm, Entergy, and Oklo.
More on META: Earnings recap