Meta Platforms Inc.
Rating
Accumulate
Adding on Dips — Active Accumulation
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
Unrivaled social graph network effect, irreplaceable proprietary data, and a rapidly growing AI platform across 3B+ users. On April 8, 2026, Meta debuted **Muse Spark** — its first frontier model from Meta Superintelligence Labs (led by Alexandr Wang, formerly Scale AI CEO), deployed natively across Facebook, Instagram, WhatsApp, Threads, and Ray-Ban glasses. Alongside the Q2 2026 print Meta shipped **Muse Spark 1.1** and finally opened the **public developer API** that had slipped repeatedly through H1, positioning it as its strongest agentic and coding model at prices below the comparable OpenAI and Anthropic tiers. Muse Spark's 'contemplating mode' (parallel multi-agent reasoning) leads benchmarks on health reasoning (HealthBench Hard: 42.8 vs GPT-5.4's 40.1) and multimodal figure understanding (CharXiv: 86.4 vs GPT-5.4's 82.8), cementing Meta's position as a full frontier AI competitor — not just an AI-enhanced advertiser. What Q2 2026 added is the price of holding that position: operating margin fell to 31% from 43% a year earlier and free cash flow to $784M, so the frontier-AI seat is now visibly paid for out of shareholder cash flow rather than out of slack. The moat is intact; the returns on defending it are the open question.
Meta's moat is built on Attention, Data, and AI Platform:
- Social Graph Network Effect: Every new user on Instagram or WhatsApp increases the value for existing users. Breaking this flywheel requires a multi-billion person migration.
- AI Content Flywheel: AI-driven recommendations are significantly increasing time-spent on Reels, which directly translates to more ad-inventory. The flywheel is now the *only* source of impression growth that matters: family daily active people compounded at just +3% YoY in Q2 2026, so the +14% impression growth came almost entirely from time-spent and ad load — a real capability, but a bounded one.
- Vertical Integration of AI: By owning the compute, the models (Llama, now Muse Spark), and the distribution (FB/IG), Meta controls the entire AI value chain.
- Meta AI Platform Moat: With Meta AI at consumer scale across 3.60B daily users and the Muse Spark 1.1 developer API live from July 2026 at below-OpenAI/Anthropic pricing, Meta now has both halves of an AI platform: the consumer memory layer and a third-party developer surface. The API is new and unquantified — no disclosed developer count or revenue — but the multi-quarter delay that was the visible execution risk has resolved, and Llama already established the open-weights standard the API inherits.
- AI Infrastructure Ownership: By owning custom AI chips (MTIA), proprietary data centers, and exploring carbon-free energy sources, Meta controls its compute destiny at a cost structure no challenger can match for its own ad/AI workloads — a real internal cost-moat that reduces reliance on AWS/Azure. The July 2026 'Meta Compute' GPU-rental business was materially walked back on the Q2 call: management said there is nowhere near enough compute for internal demand and that it would be foolish to sell capacity for short-term profit when the same silicon can be used to build intelligence that compounds. No revenue, capacity, or timing detail was given. Treat it as an option Meta has deliberately chosen not to exercise, not as a business — and note that undercutting AWS/Azure on price was never a moat anyway, since it lacks the enterprise relationships, support depth, and ecosystem lock-in that make the hyperscalers sticky.
Ten Moats Verdict
Meta's moat came through Q2 2026 unchanged and, in two places, better documented. Muse Spark 1.1 shipped with the public developer API that had slipped through H1, closing the one visible execution gap in the `learnedInterfaces` story and adding a developer-side switching cost to the consumer memory layer; Family of Apps Other revenue at +73% YoY is the first quarter in which `transactionEmbedding` shows up in the revenue line at pace rather than only in business counts. Combined with the unrivaled social graph (networkEffects: strong), the irreplaceable behavioral dataset (proprietaryData: strong), and AWS/Azure-independent compute ownership (MTIA chips, owned data centers), Meta holds 6 strong moats out of 10 applicable — unchanged from April and still the highest in its history. Two things did not change and should not be read as moat events: the collapse in operating margin to 31% and in quarterly free cash flow to $784M are the *cost* of defending this position, charged in the growth and valuation pillars where they belong. The one genuine moat negative in the print is on the regulatory side — a $2.4B charge for legal proceedings, which confirms `regulatoryLockIn` as a net cost centre rather than an advantage, alongside the structural GDPR and DMA cap. Meta Compute, announced July 1 as a route to monetising the build, was deferred on the call: management would rather consume the capacity internally, which is a defensible capital-allocation choice but removes the nearest-term evidence that the infrastructure is an asset to third parties and not only to Meta.
Muse Spark — Meta's frontier model line from Superintelligence Labs, deployed natively across all Meta surfaces — creates a genuinely strong learned interface moat, and Q2 2026 strengthened it on the one axis that was still open. Muse Spark 1.1 shipped with the **public developer API** that had slipped repeatedly through H1 2026, positioned as Meta's strongest agentic and coding model at prices below the comparable OpenAI and Anthropic tiers. On the consumer side the original thesis holds: Muse Spark is competitive with GPT-5.4 and Gemini 3.1 Pro on key benchmarks (HealthBench Hard: 42.8 vs GPT-5.4's 40.1; CharXiv: 86.4 vs 82.8), and its 'contemplating mode' and persistent memory architecture mean every interaction deepens a user's unique AI context that cannot be migrated — switching to any competing assistant means losing the accumulated memory of conversations, preferences, and context built across Facebook, Instagram, WhatsApp, Threads, and Ray-Ban glasses. The API adds a developer-side switching cost on top of that, though it is new and Meta has disclosed neither developer count nor revenue.
Advantage+ AI and the Conversions API have largely rebuilt Meta's ad targeting moat post-ATT — server-side signals, AI-modeled conversions, and closed-loop attribution now outperform the pre-ATT pixel era for many advertisers. Q2 2026 gives the cleanest read yet on the pricing half of that: average price per ad rose 12% for a second consecutive quarter and management attributed the 27% ad revenue growth to AI efficiency gains across the ads system. It stays `intact` rather than `strong` because the volume half is weakening in parallel — impressions decelerated from +19% to +14% against a +3% DAP base — so the mechanism is demonstrably valuable but is not currently widening the gap to competitors.
Open Graph API restrictions and GDPR have dismantled the public social data access advantage.
N/A — Meta's moat is network effects and proprietary social data, not scarce talent. AI tools have further reduced talent barriers for advertisers and content creators using the platform.
Meta AI with persistent memory across FB/IG/WA/Threads creates a strong cross-app bundle lock-in that didn't exist pre-2024. Users accumulating Meta AI memory, preferences, and interaction history across the entire family of apps face meaningful switching costs — the AI layer is the new bundling glue that makes the multi-app family stickier than the pre-AI era. With four billion-scale apps sharing a single AI memory layer, this bundling depth exceeds most enterprise SaaS multi-product integration.
3B+ users' social graph, behavioral patterns, and relationship data is genuinely irreplaceable — the richest consumer dataset on Earth. Family daily active people reached 3.60B in June 2026; the +3% YoY growth rate is a monetisation constraint, not a moat constraint, since the dataset's value is its depth and history rather than its rate of accretion.
Advertisers have built compliance infrastructure around Meta's Conversions API and privacy tools, creating some switching friction. However, GDPR and the EU Digital Markets Act actively constrain cross-app data sharing and limit the full expression of this lock-in, and Q2 2026 put a number on the other side of the ledger: a $2.4B charge for legal proceedings, large enough on its own to cut ~4 points off the quarter's operating margin. Regulation is a net cost centre here, not a moat.
The social graph itself IS the product — 3 billion users and their connections cannot be replicated by a new entrant in a decade.
WhatsApp Business serves 200M+ businesses globally; WhatsApp Pay is active in India and Brazil with expanding markets. Meta Pay and Marketplace create deeply embedded commerce infrastructure that is increasingly difficult to displace. Q2 2026 is the first quarter where the financials corroborate the qualitative claim at pace: Family of Apps Other revenue — predominantly paid business messaging and Meta Verified — grew 73% YoY, accelerating from +54% in Q4 2025, and is the only accelerating revenue line in the business.
WhatsApp serves as the primary communication record for billions in emerging markets and Meta AI's persistent memory is creating a nascent personal AI system-of-record, but consumer social/messaging does not carry the compliance and audit-trail migration penalties that earn 'strong' in enterprise contexts. There is no regulatory acceptance at risk, no legal chain-of-custody requirement — switching friction is real but driven primarily by network effects (already captured separately) rather than irreversible record-keeping lock-in.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
Unrivaled social graph network effect, irreplaceable proprietary data, and a rapidly growing AI platform across 3B+ users. On April 8, 2026, Meta debuted **Muse Spark** — its first frontier model from Meta Superintelligence Labs (led by Alexandr Wang, formerly Scale AI CEO), deployed natively across Facebook, Instagram, WhatsApp, Threads, and Ray-Ban glasses. Alongside the Q2 2026 print Meta shipped **Muse Spark 1.1** and finally opened the **public developer API** that had slipped repeatedly through H1, positioning it as its strongest agentic and coding model at prices below the comparable OpenAI and Anthropic tiers. Muse Spark's 'contemplating mode' (parallel multi-agent reasoning) leads benchmarks on health reasoning (HealthBench Hard: 42.8 vs GPT-5.4's 40.1) and multimodal figure understanding (CharXiv: 86.4 vs GPT-5.4's 82.8), cementing Meta's position as a full frontier AI competitor — not just an AI-enhanced advertiser. What Q2 2026 added is the price of holding that position: operating margin fell to 31% from 43% a year earlier and free cash flow to $784M, so the frontier-AI seat is now visibly paid for out of shareholder cash flow rather than out of slack. The moat is intact; the returns on defending it are the open question.
Growth Score
Q2 2026 revenue of $60.8B grew 28% YoY, beating the ~$60.2B consensus and landing near the top of the $58–61B guide — but growth peaked last quarter, not this one. The sequence now reads +24% (Q4 2025), +33% (Q1 2026), +28% (Q2 2026), and the Q3 guide of $61–64B implies roughly +22% at the midpoint, below consensus. Ad revenue grew 27% on impressions +14% (down from +19% in Q1) against an unchanged +12% average price per ad, while Family of Apps Other revenue — WhatsApp paid messaging and Meta Verified — grew 73%. Family daily active people reached 3.60B for June 2026, +3% YoY, a second consecutive quarter of deceleration; with DAP compounding at 3%, impression growth has to come from ad load and time-spent, both bounded. The cash line is where the quarter broke; the earnings line mostly did not. GAAP diluted EPS of $6.18 missed the ~$7.17–7.23 consensus and fell 13% YoY, but Q2 carried $2.4B of legal-proceedings charges and $1.18B of severance from the May 2026 reduction of ~8,000 roles — roughly $1.17 per diluted share after the 16% tax rate. Ex both, EPS was ~$7.35, up ~3% from $7.14 a year earlier and above consensus, and operating income was ~$22.4B (+9% YoY) against a GAAP print of $18.8B (−8%). Management still expects full-year operating income above 2025. What remains after stripping the one-offs is the structural cost of the build: total costs and expenses rose 55% YoY to $42B, operating margin fell to 31% from 43% (≈37% ex the charges), and capex of $31.1B — up from $17.0B a year earlier — took free cash flow to $784M from ~$8.5B on $31.9B of operating cash flow. Management narrowed FY2026 capex to $130–145B (from $125–145B, against $72.2B spent in 2025), raised the FY2026 total-expense floor to $165–169B specifically to incorporate the $2.4B legal charge, lifted the remaining-quarters tax rate to 15–17% from 13–16%, and guided FY2026 Reality Labs losses roughly in line with 2025's ~$19B (Q2: revenue $431M, operating loss $4.62B). On the other side of the ledger, Muse Spark 1.1 shipped with the long-delayed public developer API, and Zuckerberg attributed the ad strength to AI efficiency gains across the ads system. The stock fell ~8.6% to ~$535 on July 30 — as much as 10.4% intraday — and has since recovered to ~$557. This is the quarter in which the capex-ROI question stopped being hypothetical: revenue still compounds in the high twenties, and free cash flow no longer grows with it.
Valuation Score
At ~$557 (August 3, 2026) — recovered from the ~$535 post-print low, still ~10% below the mid-July ~$620 — META trades ~20% below a base fair value of $700, cut from $820 after Q2. The cut was right on the structural arithmetic and wrong on one detail: the GAAP EPS miss of $6.18 was almost entirely the $2.4B legal-proceedings charge and $1.18B of severance (~$1.17/share after tax). Ex both, Q2 EPS was ~$7.35 — a beat of the ~$7.17–7.23 consensus and +3% YoY — and management still guides full-year operating income above 2025. The FY2026 expense floor was raised to $165–169B specifically to incorporate that $2.4B legal charge; strip it, and the run-rate earnings power against ~$250B of revenue sits nearer $28–29 once the non-recurring $8.03B Q1 tax benefit is also removed, not a collapse in the advertising engine. What still justifies the cut from $820 is the cash and depreciation path the one-offs do not explain: free cash flow of $784M on $31.1B of quarterly capex, operating margin at ~37% ex charges against 43% a year ago, and 2027 capex still unguided. Revenue compounds — $60.8B at +28%, ~+22% guided for Q3, an 18–22% three-year blend — and free cash flow does not. What keeps this an Accumulate rather than a Hold is that the corridor is honest at ~$557: the price sits ~33% above a bear case that assumes the capex never pays back, the moat is unimpaired at 82, and the Muse Spark 1.1 public API and Family of Apps Other at +73% are both real optionality the multiple is not paying for.
The Advertising Moat
Meta's moat is built on Attention, Data, and AI Platform:
- Social Graph Network Effect: Every new user on Instagram or WhatsApp increases the value for existing users. Breaking this flywheel requires a multi-billion person migration.
- AI Content Flywheel: AI-driven recommendations are significantly increasing time-spent on Reels, which directly translates to more ad-inventory. The flywheel is now the *only* source of impression growth that matters: family daily active people compounded at just +3% YoY in Q2 2026, so the +14% impression growth came almost entirely from time-spent and ad load — a real capability, but a bounded one.
- Vertical Integration of AI: By owning the compute, the models (Llama, now Muse Spark), and the distribution (FB/IG), Meta controls the entire AI value chain.
- Meta AI Platform Moat: With Meta AI at consumer scale across 3.60B daily users and the Muse Spark 1.1 developer API live from July 2026 at below-OpenAI/Anthropic pricing, Meta now has both halves of an AI platform: the consumer memory layer and a third-party developer surface. The API is new and unquantified — no disclosed developer count or revenue — but the multi-quarter delay that was the visible execution risk has resolved, and Llama already established the open-weights standard the API inherits.
- AI Infrastructure Ownership: By owning custom AI chips (MTIA), proprietary data centers, and exploring carbon-free energy sources, Meta controls its compute destiny at a cost structure no challenger can match for its own ad/AI workloads — a real internal cost-moat that reduces reliance on AWS/Azure. The July 2026 'Meta Compute' GPU-rental business was materially walked back on the Q2 call: management said there is nowhere near enough compute for internal demand and that it would be foolish to sell capacity for short-term profit when the same silicon can be used to build intelligence that compounds. No revenue, capacity, or timing detail was given. Treat it as an option Meta has deliberately chosen not to exercise, not as a business — and note that undercutting AWS/Azure on price was never a moat anyway, since it lacks the enterprise relationships, support depth, and ecosystem lock-in that make the hyperscalers sticky.
Ten Moats Verdict
Meta's moat came through Q2 2026 unchanged and, in two places, better documented. Muse Spark 1.1 shipped with the public developer API that had slipped through H1, closing the one visible execution gap in the `learnedInterfaces` story and adding a developer-side switching cost to the consumer memory layer; Family of Apps Other revenue at +73% YoY is the first quarter in which `transactionEmbedding` shows up in the revenue line at pace rather than only in business counts. Combined with the unrivaled social graph (networkEffects: strong), the irreplaceable behavioral dataset (proprietaryData: strong), and AWS/Azure-independent compute ownership (MTIA chips, owned data centers), Meta holds 6 strong moats out of 10 applicable — unchanged from April and still the highest in its history. Two things did not change and should not be read as moat events: the collapse in operating margin to 31% and in quarterly free cash flow to $784M are the *cost* of defending this position, charged in the growth and valuation pillars where they belong. The one genuine moat negative in the print is on the regulatory side — a $2.4B charge for legal proceedings, which confirms `regulatoryLockIn` as a net cost centre rather than an advantage, alongside the structural GDPR and DMA cap. Meta Compute, announced July 1 as a route to monetising the build, was deferred on the call: management would rather consume the capacity internally, which is a defensible capital-allocation choice but removes the nearest-term evidence that the infrastructure is an asset to third parties and not only to Meta.
Muse Spark — Meta's frontier model line from Superintelligence Labs, deployed natively across all Meta surfaces — creates a genuinely strong learned interface moat, and Q2 2026 strengthened it on the one axis that was still open. Muse Spark 1.1 shipped with the **public developer API** that had slipped repeatedly through H1 2026, positioned as Meta's strongest agentic and coding model at prices below the comparable OpenAI and Anthropic tiers. On the consumer side the original thesis holds: Muse Spark is competitive with GPT-5.4 and Gemini 3.1 Pro on key benchmarks (HealthBench Hard: 42.8 vs GPT-5.4's 40.1; CharXiv: 86.4 vs 82.8), and its 'contemplating mode' and persistent memory architecture mean every interaction deepens a user's unique AI context that cannot be migrated — switching to any competing assistant means losing the accumulated memory of conversations, preferences, and context built across Facebook, Instagram, WhatsApp, Threads, and Ray-Ban glasses. The API adds a developer-side switching cost on top of that, though it is new and Meta has disclosed neither developer count nor revenue.
Advantage+ AI and the Conversions API have largely rebuilt Meta's ad targeting moat post-ATT — server-side signals, AI-modeled conversions, and closed-loop attribution now outperform the pre-ATT pixel era for many advertisers. Q2 2026 gives the cleanest read yet on the pricing half of that: average price per ad rose 12% for a second consecutive quarter and management attributed the 27% ad revenue growth to AI efficiency gains across the ads system. It stays `intact` rather than `strong` because the volume half is weakening in parallel — impressions decelerated from +19% to +14% against a +3% DAP base — so the mechanism is demonstrably valuable but is not currently widening the gap to competitors.
Open Graph API restrictions and GDPR have dismantled the public social data access advantage.
N/A — Meta's moat is network effects and proprietary social data, not scarce talent. AI tools have further reduced talent barriers for advertisers and content creators using the platform.
Meta AI with persistent memory across FB/IG/WA/Threads creates a strong cross-app bundle lock-in that didn't exist pre-2024. Users accumulating Meta AI memory, preferences, and interaction history across the entire family of apps face meaningful switching costs — the AI layer is the new bundling glue that makes the multi-app family stickier than the pre-AI era. With four billion-scale apps sharing a single AI memory layer, this bundling depth exceeds most enterprise SaaS multi-product integration.
3B+ users' social graph, behavioral patterns, and relationship data is genuinely irreplaceable — the richest consumer dataset on Earth. Family daily active people reached 3.60B in June 2026; the +3% YoY growth rate is a monetisation constraint, not a moat constraint, since the dataset's value is its depth and history rather than its rate of accretion.
Advertisers have built compliance infrastructure around Meta's Conversions API and privacy tools, creating some switching friction. However, GDPR and the EU Digital Markets Act actively constrain cross-app data sharing and limit the full expression of this lock-in, and Q2 2026 put a number on the other side of the ledger: a $2.4B charge for legal proceedings, large enough on its own to cut ~4 points off the quarter's operating margin. Regulation is a net cost centre here, not a moat.
The social graph itself IS the product — 3 billion users and their connections cannot be replicated by a new entrant in a decade.
WhatsApp Business serves 200M+ businesses globally; WhatsApp Pay is active in India and Brazil with expanding markets. Meta Pay and Marketplace create deeply embedded commerce infrastructure that is increasingly difficult to displace. Q2 2026 is the first quarter where the financials corroborate the qualitative claim at pace: Family of Apps Other revenue — predominantly paid business messaging and Meta Verified — grew 73% YoY, accelerating from +54% in Q4 2025, and is the only accelerating revenue line in the business.
WhatsApp serves as the primary communication record for billions in emerging markets and Meta AI's persistent memory is creating a nascent personal AI system-of-record, but consumer social/messaging does not carry the compliance and audit-trail migration penalties that earn 'strong' in enterprise contexts. There is no regulatory acceptance at risk, no legal chain-of-custody requirement — switching friction is real but driven primarily by network effects (already captured separately) rather than irreversible record-keeping lock-in.
Growth Analysis
Growth Drivers
Key Risk
FY2027 capex is unguided while FY2026 is already $130–145B against $72.2B spent in 2025, so the depreciation step-up that took Q2 operating margin to 31% has at least another year to run — and the FY2026 expense guide of $165–169B implies H2 quarterly expenses near $45B, above Q2's $42B even though Q2 contained $3.6B of one-off legal and severance charges. The falsifiable test: if operating margin prints below 30% in any quarter of 2027 while ad revenue growth falls below 15%, EPS declines outright and the compounder framing breaks — Meta becomes a capital-intensive infrastructure business that happens to own an ad network. Secondary residual risks: family DAP at +3% leaves impression growth dependent on ad load, which is bounded; the $2.4B Q2 legal-proceedings charge shows the social-media liability docket is now large enough to move a quarter on its own; and Meta Compute, the one disclosed route to third-party capex payback, was walked back on the call with no revenue, capacity, or timing detail.
Score Derivation
83.3 base − 4 margin − 5 risk = 74
Base 83 (18–22% blended revenue CAGR, decayed from the reported +28% and the ~+22% Q3 guide) + 0 trajectory (Family of Apps Other accelerating at +73%, core advertising decelerating across +33% → +28% → ~+22% guided, Reality Labs stable) − 4 margin compression (Q2 operating margin 31% vs 43% a year ago; ~37% even ex the $2.4B legal charge and $1.18B severance — the GAAP EPS miss of $6.18 vs ~$7.20 consensus was almost entirely those one-offs; ex both, EPS ~$7.35 was a beat and +3% YoY) − 5 moderate residual risk = 74. The prior derivation narrated 86 off a 20–25% CAGR, expanding margins and a high risk severity. Two of those three inputs inverted with the Q2 print. The third is cut deliberately: the risk that severity carried — capex without payback, FCF compression, a re-rate from compounder to infrastructure bet — is no longer unmaterialised. Free cash flow printed $784M, operating margin compressed to ~37% even ex the one-off charges, and the stock re-rated ~9% lower on the print. Charging that in the CAGR base and in marginTrend while also holding severity at high would charge the same fact twice. Do not also charge the GAAP EPS print — it was the one-offs.
Growth Drivers (3-Year Horizon)
Research Covering This Name
Meta's Capex Has One Receipt, and It Is Price Per Ad
Meta spent $31.1B in a quarter and printed $784M of free cash flow, and the tape traded the capex line. The line that tests the build is the other one: price per ad up 12% for a second straight quarter while impressions decelerated. Only the outcome owner can charge for a better match.
The Odds on Alphabet's $200B AI Build
Alphabet beat on revenue, on Cloud and on margin, then fell about 5% because capex went up. The market is sorting hyperscalers by how much they spend. The variable that decides whether a build pays off is how much of it was sold before it was poured.
Price Scenarios (12–24 Months)
Valuation Analysis
Using a 10-year DCF with a 9% WACC, ~12% near-term EPS growth — FY2026 run-rate EPS lands near $28–29 once the $2.4B Q2 legal charge and the non-recurring $8.03B Q1 tax benefit are stripped, as depreciation, a 15–17% tax rate and ~$19B of Reality Labs losses absorb a ~25% revenue increase — decelerating to 3.5% terminal growth, and $130–145B of 2026 capex against unguided 2027 spend, our fair value estimate for META is $700/share.
Valuation Multiples
| P/E on FY2025 EPS | ~23.7× |
| Forward P/E (FY2026E) | ~19.5× |
| Forward P/E (FY2027E) | ~17× |
| PEG Ratio | ~1.0× |
| Price / Sales (NTM) | ~5.2× |
| Price / Free Cash Flow | n/m |
The cheap-relative-to-growth case for META is weaker than it was in June, and the reason is free cash flow and structural margin — not the GAAP EPS print. At ~$557 the stock is ~10% below the pre-print ~$620; Q2 diluted EPS of $6.18 looked like a dollar miss, but ~$1.17 of that was after-tax legal and severance, and ex both the quarter earned ~$7.35. FY2026 run-rate earnings power near $28–29 (ex the Q1 tax benefit and the Q2 legal charge) against the ~$35 the June note assumed is still a real cut — depreciation, the 15–17% tax rate and Reality Labs do that work — so the multiple is ~19.5× rather than ~17.7× and PEG has gone from ~0.80× to ~1.0×. Still not expensive for 18–22% revenue growth; no longer a Mag-7 standout. The Price/FCF row is the one that actually changed character: with $130–145B of capex against roughly matching operating cash flow, FY2026 free cash flow is close to break-even and the multiple is uninformative. Reality Labs continues to cost ~$19B a year for $431M of quarterly revenue. The offsets are genuine and unpriced — Family of Apps Other at +73%, the Muse Spark 1.1 public API now live and undercutting OpenAI and Anthropic on price, and 3.60B daily users to deploy it across — but they are options, and 2027 capex is the unguided variable that decides whether they are exercised out of a growing cash flow or a shrinking one.
Approximate figures as of August 3, 2026 (Q2 2026 actuals; price ~$557).
Where We Are vs Targets
Loading live price…
The 2026–27 capex build delivers no measurable third-party revenue, ad growth decelerates through the mid-teens as impression growth converges on a +3% DAP base, and margin compression continues into 2027 so that EPS declines outright.
- Ad revenue growth falls below 15% as the impression engine runs out of ad load: DAP compounds at +3%, impressions decelerate from +14% toward single digits, and the +12% price/ad that has carried two quarters normalises in a softer ad market
- 2027 capex is guided higher again and the depreciation step-up holds operating margin below 30%; FY2027 EPS falls below FY2026's ~$27–28 and the market completes the re-rate from compounder to capital-intensive infrastructure owner at 15× earnings
- Free cash flow stays near the $784M printed in Q2 rather than recovering, constraining the $50B+ buyback pace; the $2.4B Q2 legal charge proves to be the first of a series as the social-media liability docket and EU/DMA constraints compound, and Reality Labs stays at ~$19B of annual losses for sub-$2B of revenue
Revenue compounds at 18–22% as guided, operating margin stabilises in the low thirties once the 2026 capex cohort is fully in the depreciation base, and free cash flow recovers through 2027 without the AI build having to produce third-party revenue.
- Revenue lands near $250B in FY2026 and ~$300B in FY2027: the $61–64B Q3 guide is met, ad revenue holds high-teens-to-low-twenties growth on continued +10–12% price/ad expansion, and Family of Apps Other keeps compounding above 50% toward a genuinely material revenue share
- FY2026 total expenses come in within the $165–169B guide and FY2026 EPS lands at ~$27–28 ex the Q1 tax benefit; 2027 capex growth decelerates from 2026's near-doubling, operating margin stabilises in the low thirties, and FY2027 EPS reaches ~$33 — a ~21× multiple at this target
- Free cash flow recovers to a $30B+ annual run-rate during 2027 as revenue growth outpaces the capex step-up; Reality Labs losses stay capped near $19B per the guide; the Muse Spark 1.1 API builds a developer base without yet being a reportable revenue line
The capex build converts: Muse Spark becomes a monetised platform rather than a cost centre, Meta exercises the Meta Compute option into a real third-party business, and WhatsApp commerce inflects — so margin recovers alongside 20%+ revenue growth and the multiple re-rates on demonstrated AI payback.
- The Muse Spark 1.1 API takes meaningful share of agentic and coding workloads on price, becoming a disclosed revenue line during 2027, and consumer Muse Spark lifts session time enough to push impression growth back above 20% despite the +3% DAP base
- Meta exercises the Meta Compute option it deferred in July 2026, selling excess capacity into a compute-scarce market at attractive incremental margins — turning the $130–145B build from a cost line into an asset with a visible return and resetting how the market capitalises the capex
- WhatsApp commerce and business messaging inflect to $15B+ by 2027 on the +73% Family of Apps Other trajectory; operating margin recovers toward the high thirties, FY2027 EPS reaches ~$35, and the multiple re-rates to ~25× on confirmed AI payback