Built by Grok 4.6 High

Meta Platforms reported third-quarter 2026 results on Wednesday, October 21, 2026. This review was prepared Monday, October 26, 2026. Quarter ended September 30, 2026. Calendar fiscal year. Next hard print: Q4 2026 on January 27, 2027.

META | Earnings Review — Q3 2026

Advertising is the company: $63,061 million of $64,601 million, about 98% of sales. Eight-quarter revenue growth ran +20.6%, +16.1%, +21.6%, +26.2%, +23.8%, +33.1%, +28.0%, +26.1%. The Q3 stack itself is +18.9% then +26.2% then +26.1%. A 26% grower that has stopped speeding up is a different setup from the +33% print two quarters ago. Family daily active people were 3.64 billion, only +2.8% YoY. This is a price story, not a user story.

YoY rates are the same quarter a year earlier, calculated from the cited dollars. Margin basis points use the unrounded ratio. Q3 2025 EPS of $1.05 is the one-time tax quarter (provision $18,954 million). Q1 2026 EPS of $10.44 includes a tax benefit of $5,021 million. Free cash flow before Q3 2025 is omitted rather than filled in. Gold outline is this quarter.

The United States is still accelerating. Europe and Asia-Pacific are not. U.S. and Canada advertising was $28,438 million, +33.3%, with price +23% and impressions +8%. Europe was $14,234 million, +17.9%, price +10%. Asia-Pacific was $11,531 million, +15.1%, with impressions +14% versus +23% a year ago and price stuck at +1%. Rest of world was $8,858 million, +33.0%. The consolidated slowdown is Europe and Asia-Pacific, not a U.S. demand break.

Family of Apps other revenue $1,097 million grew +59% and is 1.7% of sales. Reality Labs revenue $443 million was −5.7%. The operating loss narrowed to $4,129 million from $4,432 million. The profit problem is Family of Apps operating income, $17,117 million versus $24,967 million, −31.4%. That is where the legal charge sits.

FY2022 was the air pocket. FY2023 and FY2024 restored the ad grower and expanded EBITDA margin. FY2025 still edged revenue growth higher, to +22.2% from +21.9%, and gross margin still expanded 33 bp to 82.0%. The profit rate did not: EBITDA margin −90 bp and operating margin −74 bp. Nine-month 2026 revenue is $181.7 billion, +28.8% versus the same three quarters of 2025. A full year can still grow faster than FY2025 because the first half did. That is not the same thing as the business accelerating into year-end. FMP's FY2026 revenue average of $256.3 billion is nine-month actuals plus a Q4 near the guide midpoint, not a bet that +26% re-accelerates. FMP FY2026 EPS of $29.44 is stale against a $3.83 quarter.

Green is a beat versus the street bar used in the score. Red is a miss. Revenue actuals are GAAP. EPS actuals are GAAP diluted EPS except Q3 2025, which is scored on management's published ex-charge $7.25 because the FMP estimate was not a GAAP-$1.05 bar. This quarter's FMP epsActual of 7.40 is not scored. Gold outline is Q3 2026.

Pattern: Mixed. Revenue is a consistent beater — L4Q 4/4, L8Q 8/8, and 12/12 on the history used here. EPS has missed two quarters in a row. L4Q EPS beat rate is 50%. L12Q is 10/12. EPS surprise magnitude is deteriorating: clean beats faded from about +$1.20 in early 2025 to +$0.53–$0.69 in the second half, the Q1 GAAP blowout was a tax benefit (clean EPS would have been $7.31, still a beat of about +$0.61), and the next two gaps widened from −$1.01 to −$2.71. Revenue surprise compressed from a +6.0% peak in Q2 2025 to +1.0% in Q2 2026, then widened modestly to +2.0% and, more importantly, finished above the company's own high.

This quarter: revenue $64,601 million versus FMP $63,343 million, +$1,258 million / +2.0%, and +$601 million above the $64 billion guide high. GAAP EPS $3.83 versus $6.54, −$2.71 / −41.4%. The $6.54 bar did not reserve the $9.89 billion August multistate attorneys general charge. The tax rate was 24%, above the 15–17% outlook, because part of the legal expense is nondeductible. A full tax shield on the charge is an illustration near $6.75, not a company EPS, and it still leaves operating margin down about 470 bp. The revenue beat is high quality: price and impressions, not a one-time. The EPS miss is a poor read on ad demand and a fair read on cost.

* Q3 2025 scored on ex-charge $7.25 versus FMP $6.72. GAAP was $1.05. † Q1 2026 GAAP beat is inflated by the tax benefit. Counted as a beat either way.

Meta does not guide EPS, gross margin, operating margin, or EBITDA. The numeric guides are Q4 revenue, the Q4 tax rate, FY2026 expenses, FY2026 capex including finance-lease principal, and — new this call, transcript only — Reality Labs losses. There was no prior company guide for Q4 itself. The Q2 range of $61–64 billion was the Q3 guide.

Q4 revenue — first guide, then the street

The street is a rounding error above the midpoint. Given the beat record below, the debate is the top of the range, not whether $74.9 billion is reachable. Last year's Q4 guide asked for a much larger slowdown than the company then printed. This year's range is close to the recent run-rate, so last year's growth-rate beat should not be pasted onto $74.5 billion.

Expenses — the entire raise is the charge

Read the red bar as cost, not as a better outlook. January's expense range was already $162–169 billion. Strip the Q2 and Q3 legal charges and the outlook is still that January outlook.

Capex — floor up, ceiling stuck

Nine-month capex including finance-lease principal is $19.84 billion + $31.08 billion + $34.91 billion = $85.8 billion. The midpoint leaves about $54 billion for one quarter. Q3 free cash flow was already negative. FY2025 capex on the same definition was $72.22 billion, so the new midpoint is about +94% YoY.

They are not a guidance misser on revenue. Five of the last seven quarters cleared the high end. Q1 and Q2 2026 finished under the high by less than $0.2 billion. Q3 2026 went back through the top: actual $64,601 million versus a $61–64 billion range, +3.4% versus the midpoint. The weak record is the cost side. Capex floors have moved up at every print this year. The Q3 tax rate of 24% missed the 15–17% outlook and was attributed to the nondeductible legal portion. The Q4 tax range of 15%–17% was simply reinstated.

Implied full-year revenue, nine-month actuals plus the Q4 range, is $254.7–257.7 billion, midpoint $256.2 billion, +27.5% versus FY2025 $200,966 million. FMP's FY2026 revenue average is $256.275 billion. In line. That full-year rate is faster than FY2025 only because the first nine months already grew 28.8%.

Revenue growth troughed at +16.1% in Q1 2025, peaked at +33.1% in Q1 2026, and has given back about 700 bp of that year-over-year rate in two quarters. The second step, −189 bp, is smaller than the first, −512 bp. Constant currency confirms the latest step is organic: reported and constant-currency growth were both 26% in Q3, after constant currency had already cooled from 29% to 27%. Part of the ugly −512 bp on the reported line from Q1 to Q2 was the FX tailwind fading. What is left is a genuine, slower fade.

Since the Q1 peak, impression growth has decelerated 900 bp (+19% to +14% to +10%) while price per ad has re-accelerated 200 bp (+12% to +12% to +14%). Revenue YoY fell by less than impressions because price offset part of the volume slowdown. Susan Li tied the +10% impressions to engagement, users, and ad load, healthy in every region, and the +14% price to ad-performance gains plus a better macro than Q3 2025, partly offset by lower-monetizing surfaces. Instagram time spent and Facebook Reels time spent were described as all-time highs and up double digits. The impression rate can decelerate while time spent is still growing, because last year's impression prints of +14%, +18%, and +19% are the lap.

The Q4 guide does not reverse it. Midpoint +24.4% is about −170 bp versus Q3. Only the top of the range, +26.9%, would re-accelerate. With a 1% FX headwind, constant-currency growth at the midpoint is still a touch slower than Q3's 26% constant-currency rate.

The quarter resolved the ads debate in management's favor on the level — advertising still +26% — and left the new debates unmodeled. Consensus is underwriting a mid-guide holiday quarter. It is not underwriting Muse.

Management's contrarian claim is that Muse, at a billion users, with a commerce take-rate and a greater-than-10x efficiency gain, becomes a second earnings engine, and that the same compute also keeps the ads system growing. The street's Q4 and FY2026 numbers do not underwrite that claim. What would close the gap is a Q4 print at the high end — +26.9%, an acceleration versus last Q4's +23.8% — and a January guide that numbers 2027 capex without walking back "strong" 2027 revenue growth. What would confirm the caution is Europe price staying at +10% while U.S. price rolls over, Q4 revenue at the $73 billion low end, a January capex ceiling above $145 billion, or another legal charge.

The hour was not about the print. Revenue growth held at 26%, the same rate as Q3 2025, while GAAP operating income fell and free cash flow swung negative. The street spent the call underwriting Muse. Product questions were answered. Cash questions were not. Tally: 10 well answered, 3 deflected.

What was answered, and should not be blurred into the deflections: Muse's growth is capacity- and efficiency-gated, not feature-gated. Enterprise is the model API, Muse code, and business agents first, with revenue over years rather than in 2027. "Strong growth" in 2027 is still the ads business; new AI revenue can begin, it does not carry the year. Monetization starts sooner than the old surface playbook, because the cost and the commerce use are both large enough that they cannot scale the network without it. The commerce stat includes travel. Zuckerberg was explicit that they will not charge everyone up front.

The three deflections are one argument.

1. Unit cost (Sandler). Li refused a per-user cost and did not restate the "more than 10x" efficiency line when asked for the curve. The levers — off-peak capacity, smaller models, higher GPU utilization, fewer tokens — are a roadmap. They are not a checkable slope. Without a starting cost, 10x cannot be audited, and Zuckerberg later said the capability bar will rise as unit cost falls.
2. 2027 anchors (Anmuth). Waiting until January for a dollar guide is the standing cadence. What was avoided is whether capex dollar growth stays in the "notably larger" regime, whether 2027 is still a maximize-compute year, whether flexibility really starts in 2028, and whether operating income is still supposed to grow every year. She retired none of those anchors and reaffirmed none of them.
3. The new share class (Sheridan). The class is a non-voting currency for capital raising, compensation, acquisitions, and general purposes, without giving up Class A and Class B votes. Then: no specific financing plans. That was asked into a quarter with free cash flow of −$2,120 million and capex including finance leases of $34.91 billion. Cash $11,222 million plus marketable securities $71,227 million is about $82.4 billion, in the neighborhood of long-term debt face $84,000 million. Creating a funding tool and declining to name the marginal source is the dodge.

Nobody asked about the legal charge, the capex-floor raise, the $15 billion 2027 Reality Labs loss, or the headcount drop to "over 66 thousand, down 12% from Q2" after the May reduction. Those are gaps in what the street chose to ask, not scored deflections. The −12% is the company's sequential comment, not a year-over-year growth rate.

Two items, both high. "No contradictions found" would be false. The nearest misses — Reality Labs "similar" versus a modest decline, a narrowed capex range whose floor rose, negative quarterly free cash flow against still-positive year-to-date cash generation, and the disclosed 24% tax miss — can both be true and are not counted.

No rate, CPI, GDP, or confidence index was cited. The demand evidence is the ad auction, and the mix inside it changed. Price is carrying growth. Macro was called better than the year-ago quarter for a third straight call, even though Q3 2025 itself did not credit macro. FX has faded. The new equity read-throughs are agentic-commerce partners, plus Threads overtaking X.

Meta is still the ads leader, and the leader is still growing at +26%. That is not the argument. The argument is the second derivative. Revenue growth has decelerated for two quarters off a +33.1% peak, impression growth has cooled 900 bp, and gross margin has broken by 273 bp because infrastructure cost is outrunning revenue. The $9.89 billion legal charge makes GAAP profit look worse than the business and does not repair the margin. The Q4 guide is in line with the street and another small step down. The operating-income promise that was supposed to govern the AI spend no longer holds on a GAAP basis inside the company's own ranges, and the questions that would make the Muse build fundable — unit cost, 2027 spend, and the use of the new share class — were the ones that were deflected.

Hold the franchise. Do not underwrite a re-acceleration, and do not book FMP's 7.40 as the quarter. The next check is whether Q4 price can keep offsetting a slower impression rate, and whether the January call will say 2027 operating income can grow at the same time as capex.