Revenue Grew 26%, Clean EPS Grew 2%, and Meta Built a Non-Voting Share for the Funding Gap

Results vs. Consensus

Q3 2026 Scorecard

Quality of the beat. The revenue beat is as clean as Meta prints: constant-currency growth matched reported growth at 26%, so currency contributed nothing, and the upside came from price rather than volume. The EPS beat is real but less durable than its size. A 1.9% revenue beat produced a 13% clean EPS beat because costs came in well under models. Headcount fell by 8,697 in the quarter and R&D fell 7.4% sequentially as the May reduction cleared the payroll. That is a one-time step down in the compensation base. Depreciation, up 44% year over year and now visibly compressing gross margin, is moving the other way and is not one-time.

Year-over-Year Comparison

The prior-year column needs its own caveat. Q3 2025 carried a $15.93B one-time, non-cash tax charge tied to the One Big Beautiful Bill Act, which took its effective tax rate to 87% and GAAP EPS to $1.05. The comparable figure Meta published at the time was $7.25. Against that, this quarter's $7.40 is the honest comparison, and it is the most important number in the release.

Sequential Comparison

Q2 charges in the ex-charges row are the $2.40B legal accrual and $1.18B of severance we excluded last quarter. On that basis, operating income grew 2.3% sequentially on 6.2% revenue growth, and the clean margin slipped another 1.4 points. The year-over-year compression is narrowing (6.3 points in Q2, 4.7 points in Q3), but only because the comparison base is softening. The sequential trend has not turned.

Revenue

This is the part of the quarter that forces us to revise our view upward. In July the Q3 guide of $61–64B implied 22% growth at its midpoint, the second consecutive five-to-six-point deceleration, and we built our downgrade partly on the assumption that the roll-over would continue. It largely did not. Revenue grew 26.1%, down from 28.0% in Q2 but four points above the guide midpoint, with no help from currency, and the Q4 guide implies 21.9% to 26.9%. Growth is now easing gently in the mid-20s rather than falling toward the high teens. Geographically the strength was concentrated where Meta monetizes best: revenue by user geography grew 33% in the U.S. and Canada and 34% in Rest of World, against 18% in Europe and 16% in Asia-Pacific. The mix of drivers shifted again, with impressions slowing to 10% and price per ad accelerating to 14%, which we address in Key Topics.

Margins

Last quarter we argued the margin damage was operating expense first and depreciation later. This quarter the second wave arrived on schedule. Cost of revenue, the line where server and data-center depreciation lands, grew 45.2% and took gross margin from 82.0% to 79.3%, a 273bps decline after only 76bps in Q2. Depreciation and amortization reached $7,154M, up 44%. At the same time the first wave partly receded: R&D fell 7.4% sequentially as roughly 8,700 roles left the payroll. The two moves largely offset, which is why the clean operating margin of 35.4% looks only modestly worse than Q2's 36.8%. The composition, though, has moved from the line management can cut to the line it cannot.

EPS

GAAP EPS of $3.83 includes the $9.89B charge, which costs $3.57 per share after tax according to the company's own reconciliation, so the underlying figure is $7.40. Two below-the-line items worked against it. Interest and other income fell to $13M from $1,128M, consistent with lower cash balances and higher interest expense, a swing worth about $0.36 per share at a 17% tax rate. The clean effective tax rate of 17% sat at the top of the 15–17% guide and above the prior year's 14%. Without those two items clean EPS would have grown about 11%. With them, it grew 2.1%. Diluted shares were flat at 2,573M with zero repurchases, but basic shares rose 1.4%, and net share settlement of vested equity, which cost $4,064M of withholding taxes this quarter, is now the main brake on dilution.

The Cost Line

Last quarter the cost story was R&D. This quarter it is infrastructure, and the shift is visible in where the dollars went.

"Year over year growth was primarily driven by legal related costs infrastructure costs, and third party AI token costs. The growth in infrastructure costs was driven by higher third party cloud spend depreciation, and data center operating costs." — Susan Li, CFO

Excluding charges, cost of revenue accounted for 38% of the $11.0B cost increase, up from about 25% on the same basis in Q2. R&D accounted for 45%. The 10-Q now lists infrastructure ahead of employee compensation as the primary driver of R&D growth. G&A excluding the AG charge still rose 41% to $4,967M, and the 10-Q attributes the G&A increase mostly to legal-related costs, so other legal accruals are likely running through this line as well.

Assessment: The quarter validated the sequencing we laid out in July: opex first, depreciation second. What we got wrong was the size of the opex response. Headcount fell by nearly 8,700 in a quarter and R&D fell sequentially, which bought one quarter of margin stability. It cannot be repeated on the same scale, while depreciation will keep rising with property and equipment up $76.8B in nine months and Q4 capex implied at $49–59B. The clean margin is stable for the wrong reason.

Segment Performance

Segment Results — Q3 2026

Revenue by Geography — Q3 2026

The table uses the 10-Q's customer-address basis. On the user-geography basis Meta uses in its commentary, which assigns revenue to where ads are seen, growth was 33% in the U.S. and Canada, 18% in Europe, 16% in Asia-Pacific and 34% in Rest of World.

Family of Apps

Advertising: Price Did the Work

Advertising revenue of $63,061M grew 25.9%, and the decomposition flipped from the first half. Impressions grew 10%, the slowest rate of the year after 19% in Q1 and 14% in Q2, while average price per ad grew 14%, up from 12% in each of the first two quarters. The 10-Q attributes the price gain to "an increase in advertising demand, which we believe is mostly due to ongoing improvements to our ad performance from our ad targeting and measurement tools." Impression growth skewed to lower-monetizing regions and Reels partly offset it. The CFO also credited easier macro comparisons against Q3 2025.

"In Q3, we launched this approach to generative retrieval on Facebook feed, driving a nearly 2% increase conversions across targeted segments. In addition, we expanded our JEM model for ads ranking to additional surfaces, driving a 1.7% increase in conversions for off-site conversion campaigns and a 9.6% increase in app ad conversions." — Susan Li, CFO

Partnership ads, where advertisers run campaigns with creators, grew more than 97% and passed $15B of annual revenue run rate. Nearly 10 million advertisers used at least one generative creative tool in the quarter, and advertisers who adopted the AI Business Assistant's recommendations drove almost 5% more conversions at the same cost per conversion.

Assessment: Price-led growth is the higher-quality kind, because it means advertisers are paying more for outcomes rather than Meta simply showing more ads. It is also the kind that depends most directly on the AI spend continuing to deliver measurable conversion gains, since the impression lever is visibly tiring. For now the gains are arriving on schedule and the advertising line is the strongest part of the investment case.

Other Revenue: WhatsApp and Meta One

Other revenue grew 59.0% to $1,097M, driven by WhatsApp paid messaging and subscriptions. Growth slowed from 72.7% in Q2. Meta One, the consumer and business subscription launched in Q2, expanded globally in September and has passed 15 million subscriptions and trials.

Assessment: The line is still 1.7% of revenue and the deceleration is a reminder that subscription revenue scales more slowly than the narrative around it. The 15 million figure blends trials with paying subscribers, so it is not yet a revenue disclosure. Muse subscriptions will eventually land here, which makes this the line to watch for the first evidence that consumer AI pays.

Reality Labs

Revenue fell 5.7% to $443M as lower Quest headset sales outweighed AI glasses growth, and the operating loss narrowed to $4,129M from $4,432M a year ago and from $4,619M in Q2. The 10-Q attributes the improvement to lower employee compensation, partly offset by rising third-party AI token costs and inventory. For the first time this year management gave the segment a numeric range: a 2026 loss of $18–19B, against $19,193M in 2025, and about $15B in 2027.

"So with these factors combined, we expect that the P&L of Reality Labs to improve from here with operating loss coming down to approximately $15 billion next year and trending towards profitability from there." — Mark Zuckerberg, CEO

The arithmetic is less dramatic than the framing. The nine-month loss is $12,775M, so the full-year guide implies a Q4 loss of $5.2–6.2B, broadly in line with Q4 2025's $6,021M. The 2027 figure is where the change sits: roughly $3–4B less loss than 2026, funded by pacing VR hardware and redirecting investment "primarily" to AI.

Assessment: Last quarter we moved this risk to materializing because the loss was widening without a numeric guide. Both conditions reversed. A $15B loss is still a large number and some of the saving is a reallocation to AI spending that shows up elsewhere in the P&L, so this is not a free $4B of consolidated operating income. But it is the first concrete capital-discipline commitment Meta has made on this segment in years.

Key KPIs

Cash Flow, Balance Sheet and Capital Returns

Last quarter free cash flow nearly reached zero. This quarter it went through it.

Operating cash flow grew only 9.3%, against 24.7% in Q2, and even that figure is flattered: the $9.89B charge is an accrual paid in installments over ten years, so it reduced net income without reducing cash. Capex including finance-lease principal was $34,906M, 54.0% of revenue and 106.5% of operating cash flow. The $1,361M dividend and $4,064M of equity-withholding taxes were paid from the balance sheet.

Assessment: Three numbers in this section matter more than the free cash flow line. First, Meta is now in a small net debt position for the first time in this investment cycle, $1.2B on the CFO's own $83.7B debt figure. Second, $10.8B of money market funds now sit in escrow under multi-year infrastructure purchase agreements, excluded from the cash figure and restricted until 2028–2030, which suggests suppliers now require collateral from Meta. Third, non-cancelable commitments rose $105.2B in a single quarter to $454.5B, nearly double March's $237.7B, with $121.4B due in 2027 alone. Management declined to give 2027 capex. The 10-Q has effectively given the floor on 2027 infrastructure outflows instead.

Key Topics & Management Commentary

Overall Management Tone: The most product-confident call in a year, built around a launch with adoption numbers attached rather than a roadmap. On the financial side management stayed guarded: no Muse economics, no 2027 framework, no comment on a $25B court judgment issued hours earlier, and a share-class announcement framed as prudence rather than as a funding plan. Tone moved from expansive but defensive in Q2 to assured on product and noncommittal on cost.

1. Muse Shipped, and the Early Numbers Are Real

In July the CEO said of personal agents, "we haven't done that yet." In the three months since, Meta launched the Muse personal agent, shipped Muse Wonder as its first frontier model, and put adoption figures on the table. More than 10 million people in the U.S. and Canada set up an agent in about six weeks, roughly a quarter of them use it on a given day, and shopping is the leading use case, engaged in by more than 25% of users with an average purchase above $200.

"About 25% of people who have set up Muse use it on any given day. For our earliest adopters, the longer that they use Muse, the better the retention seems to get. So to put this in perspective, this is more than 2x the level that Threads started at or where we believe that other AI apps started at as well." — Mark Zuckerberg, CEO

The launch came with secondary evidence. Payment and data-connection partners reported unprecedented volumes, credit limits had to be raised in the first week, and engagement per person in Meta AI more than doubled since April. Muse Wonder sits in the paid tier, with a cheaper Muse Spark as the free default.

Assessment: This resolves the single largest product gap in our thesis. The training portion of the capex now has a shipped consumer product with measurable retention, in two markets, in six weeks. What it does not yet have is revenue, and the commerce data point is a usage statistic rather than a monetization one. Meta disclosed neither purchase frequency nor any take rate.

2. Muse Is Expensive to Run, and the 10x Efficiency Promise Carries the Economics

The launch's constraint is cost. Management confirmed that Muse has not launched globally partly because it is compute-intensive, and that serving costs will rise as it scales. The economic case rests on a single forward commitment: a more than tenfold improvement in compute cost efficiency within a year, through off-peak capacity, smaller distilled models and inference optimization.

"In the near term, Muse is very compute intensive to run. But we expect to improve its cost efficiency by more than 10x in the next year, which will enable it to be profitable once the revenue lines are at scale." — Mark Zuckerberg, CEO

"Muse's computing intensity means that expanding its user base will require significant investment, and we expect total serving costs to rise." — Susan Li, CFO

One data point supports the efficiency claim. Moving content understanding onto first-party Muse models reduced per-unit processing costs by more than 80% across Instagram, Threads, Facebook and ads, while outperforming the prior production models in offline evaluations.

Assessment: Model costs falling an order of magnitude per year is consistent with the industry's recent history, so the 10x target is plausible. The catch is the one the CEO named himself later in the call: users will not want last year's intelligence at last year's price, so efficiency gains get reinvested in capability. Serving cost per user may fall while total serving cost rises, and investors should model the second.

3. Monetization Moves Up the Schedule

Meta's historic playbook is to scale engagement for years before monetizing. Management said Muse will break that pattern because the compute bill is too large to defer revenue. The model has four parts: free for most users with a transaction fee paid by businesses on commerce Meta facilitates, consumer subscriptions for frontier models and extra tokens, business subscriptions, and links into Meta's ads and content personalization.

"For most people, the product should be free, and we will build commerce tools so we get paid a fee based on the transactions that we help facilitate. We expect this will be paid for by businesses." — Mark Zuckerberg, CEO

The CEO anchored ambition to the Family of Apps, where existing Muse users generate "several hundreds of dollars annually" of revenue per American user, and said Muse should eventually exceed that. Meta also previewed a personal agent protocol with partners including Walmart, Stripe and Shopify to standardize agent-to-business transactions.

Assessment: A transaction-fee model paid by merchants is the right structure for Meta, because it extends the advertiser relationship rather than asking consumers to pay. It also places Meta in direct negotiation with retailers and marketplaces that may prefer agents not to intermediate their customers. Treat 2027 revenue from Muse as small and the 2028 revenue as the real test.

4. The Class C Recapitalization: A Funding Currency That Protects the Vote

The release's subtitle announced something other than the quarter: a new class of listed, non-voting Class C stock, effective January 19, 2027, distributed as a one-for-one dividend on every Class A and Class B share, which works like a two-for-one split. The board approved it on a special committee's recommendation, and it took effect by written consent of Mark Zuckerberg's entities, which hold about 56.1% of the vote on roughly 12% of the economics. The stated purposes come first in the release: capital raising, employee compensation and acquisitions.

"We do not have any specific financing plans to announce at this time. But we thought it was prudent to establish the appropriate capital structure framework to ensure that we have access to a diverse pool of funding sources." — Susan Li, CFO

The accompanying Founder Alignment Agreement adds two features investors should read closely. Most of the founder's Class C shares can be sold only alongside voting shares, with 2.5% released each quarter he remains in a full-time leadership role, so liquidity is tied to tenure. And the board has agreed to declare another one-for-one Class C dividend each time the Class C price doubles from the September 10 reference price of $580.60 (adjusted for the dividend), provided annualized shareholder return since then is at least 10%.

Assessment: Every share Meta issues from 2027 onward, whether to fund capex, pay engineers or buy a company, can now be issued without diluting the founder's control. That is the point. It removes the governance cost of equity issuance for the controlling holder, which lowers the internal hurdle for using equity at exactly the moment free cash flow turned negative. The split mechanics are economically neutral. The financing optionality is not, and the market is right to price the possibility of dilution.

5. Free Cash Flow Turned Negative and Commitments Jumped $105B

Free cash flow of ($2,120M) compares with $10,625M a year ago and $784M in Q2. The full-year capex guide now has a $135B floor, and with $85,824M spent through September, Q4 capex must land between $49.2B and $59.2B, 41% to 70% above Q3. On any reasonable Q4 operating cash flow, Q4 free cash flow is deeply negative and full-year free cash flow lands near zero or below.

Management did not discuss the contractual commitment line, which may be the most important disclosure of the quarter. Non-cancelable commitments, mostly third-party cloud capacity, servers, network and energy agreements, reached $454.5B, up from $349.3B in June and $237.7B in March. The portion due in 2027 rose to $121.4B from $81.7B a quarter earlier.

"But we expect to continue investing aggressively to support both our core business and emerging AI initiatives. Our current capacity and what is scheduled to come online gives us the flexibility to allocate resources across model training, core business growth, and the development and support of new products." — Susan Li, CFO

Assessment: The phrase "continue investing aggressively" plus $121.4B of 2027 commitments is a 2027 capital budget in all but name, and it points to the upper half of the $150–185B range we carried in July, which is why we raise our 2027 range below. Meta can fund it: $82.4B of liquidity, $96.9B of nine-month operating cash flow and open debt markets are not a constraint. The question is no longer whether Meta can pay. It is what the shareholder receives, and for 2026 the answer is a dividend paid from the balance sheet and no buyback.

6. The Legal Bill: $9.89B Charged, $18.59B Agreed, $25B Judged

In Q1 we called youth-safety litigation the most underpriced overhang on the stock. Q2 brought a $2.4B accrual. Q3 brought the settlement: an agreement with 52 attorneys general for payments of up to about $18.59B over ten years. About $12.7B is fixed and payable in annual installments. The remaining $5.3B is contingent on peers such as YouTube and TikTok adopting similar terms, is not accrued, and could be expensed later. The agreement also requires product changes for U.S. users under 18 and an independent compliance auditor for five years.

On the day of the print a New Mexico court imposed a separate $25B civil penalty, with 15% post-judgment interest, after a September jury verdict covering more than 43 million violations of the state's Unfair Practices Act. Meta says it has accrued an amount and will appeal. A consolidated personal-injury bellwether trial begins October 28.

"Additionally, we continue to monitor active legal and regulatory matters that could significantly impact our business and financial results. For example, we continue to see scrutiny on youth related issues in several markets, and have a number of youth related trials scheduled in the coming months in the United States." — Susan Li, CFO

The Q2 version of that sentence ended with "which may ultimately result in a material loss." This quarter it did not, which is consistent with the multistate settlement having absorbed most of the exposure that language described.

Assessment: The settlement is expensive and good news at the same time: it converts an unbounded multistate risk into a schedule of roughly $1.3B a year. The New Mexico judgment is the opposite, a single-state penalty large enough that Meta warns it may need to post a bond on appeal, in a case the settlement excluded. We do not treat the $25B as a likely outcome, but we haircut fair value for the unaccrued portion of both items, and we expect further charges rather than a clean docket.

7. Price Carried the Ad Business as Impressions Slowed

Impression growth has slowed in each of the last two quarters, from 19% in Q1 to 14% and now 10%, the slowest rate in at least five quarters. Management attributed impression growth to engagement, users and ad load, with strength in lower-monetizing regions. Pricing accelerated to 14%, the fastest rate in at least five quarters and up from 6% in Q4 2025.

"The global average price per ad increased 14% year over year driven by ad performance gains and improvements in macro conditions relative to Q3 of last year. This was partially offset by strong impression growth particularly from lower monetizing surfaces and regions." — Susan Li, CFO

Engagement disclosures support the price story: Instagram time spent per user reached an all-time high and grew double digits, Facebook Reels time grew double digits globally and in the U.S., and Threads time per user grew more than 20%. Relevance launches lifted Instagram Reels time spent by more than 4%, and nearly 45% of recommended Reels are now less than a day old.

Assessment: As long as price per ad grows faster than impressions decelerate, the revenue line holds in the mid-20s. The risk is concentrated: the price gain depends on conversion improvements, and part of it was an easy macro comparison that fades in 2027. We do not think the impression slowdown signals a user problem, since DAP and time spent are both rising.

8. The Depreciation Wave Reached Gross Margin

Gross margin of 79.3% was the first reading below 80% in at least five quarters and the largest year-over-year decline of 2026, after 26bps in Q1 and 76bps in Q2. Cost of revenue grew 45.2%, which the 10-Q attributes "primarily" to data-center, technical infrastructure and third-party cloud costs. Depreciation and amortization reached $7,154M, up 44% year over year, against nine-month property and equipment growth of $76.8B.

Management did not mention MTIA or custom silicon on the call for a second consecutive quarter. The only cost-efficiency evidence offered was model-level, such as the 80% reduction in content-processing cost from moving to first-party models.

Assessment: We argued in July that the depreciation step-down in gross margin had barely begun. It has begun. With Q4 capex implied at more than $49B and 2027 commitments at $121.4B, depreciation growth accelerates from here, and the in-house silicon argument that was supposed to offset it has gone silent. This bull pillar no longer holds as written.

9. Headcount Down 15%, R&D Down Sequentially

Headcount ended at 66,775, down 15% year over year and 8,697 from June, as the May reduction cleared. Management had said the majority of the roughly 8,000 affected employees would leave by the end of Q3, and the decline was larger than that. R&D fell to $20,054M from $21,656M in Q2, although Q2 costs included $1.18B of severance.

"We ended Q3 with over 66 thousand employees, down 12% from Q2 following the May 2026 headcount reduction." — Susan Li, CFO

Assessment: This delivered on a Q2 commitment and is the main reason the clean margin held near Q2's level. If all of Q2's severance sat in R&D, as our Q2 analysis assumed, R&D fell by only about $0.4B sequentially, so most of the payroll saving was reinvested in infrastructure and token costs inside the same line. The cut changed the mix of R&D more than its size.

10. Reality Labs Gets a Loss Guide and a Smaller VR Budget

At Connect the company said it will ship new VR devices only when there is a major breakthrough, and Reality Labs is now organized around putting Muse on glasses, starting next month. The CEO described glasses as priced for platform adoption rather than hardware profit, with subscriptions as the business model.

"I also shared at Connect that we are changing our approach to virtual reality to only ship new devices when we have major breakthroughs like we had with the Meta VR glasses this year, which, of course, are the first VR experience ever in a pair of glasses." — Mark Zuckerberg, CEO

Assessment: Pacing VR is the decision bears have asked for since 2022, and it comes with numbers: $18–19B this year and about $15B next. The metaverse spending that defined Reality Labs is being subordinated to AI, which is a better use of the capital even if the consolidated savings are smaller than the segment savings.

11. Enterprise and the Fixed-Cost Logic of Owning Frontier Models

Meta formalized its enterprise effort as the Meta Enterprise platform, led by CJ Desai, offering the model API, Muse Code and business agents. The CEO framed it as a utilization argument: training is a large fixed cost, so once Meta has leading models it should sell as much high-margin inference as it can.

"1 other note on compute economics is that since training is a large fixed cost, now that we have leading models, we want to serve as much high margin inference demand as possible. That is why we are building the Meta Enterprise platform led by CJ Desai." — Mark Zuckerberg, CEO

The CFO tied the opportunity to the 500 million self-identified small businesses already on Meta's apps and said Meta will scale its investment "behind demonstrated customer demand and product market fit in this area."

Assessment: Selling intelligence rather than raw compute is the higher-margin version of the compute-rental option we flagged last quarter, and naming a leader is progress on the go-to-market gap management admitted in July. The prior quarter's claim of premium offers for Meta's compute was not repeated. Until the enterprise line has a revenue figure, it is optionality we do not capitalize.

Guidance & Outlook

The revenue guide is the strongest part of the outlook and the reason the stock rose after hours. At $73–76B, Q4 implies year-over-year growth of 21.9% to 26.9% against Q4 2025's $59,893M, with a 24.4% midpoint, and the guide assumes a 1% currency headwind, so underlying growth runs about a point above the reported range. Sequentially the midpoint is a 15.3% step up from Q3, slightly below the 16.9% Q3-to-Q4 gain of 2025, so the guide is not stretching.

Implied Q4 margin: The expense guide of $175–179B less $127,077M spent through September implies Q4 expenses of $47.9–51.9B. At both midpoints, Q4 operating income is about $24.6B, a 33.0% margin against 41.3% in Q4 2025. That is an 8.3-point year-over-year compression, larger than the 4.7 points in Q3 on a clean basis. Even at the high end of revenue and the low end of expenses, the Q4 margin is about 37%, still below last year.

What the operating income commitment now requires. Management now holds the "above 2025" commitment only excluding the Q3 charge. Through September, operating income is $54,635M, or $64,525M with the charge added back, so Q4 needs only $18.8B to clear 2025's $83,276M. At guide midpoints, ex-charge full-year operating income is about $89.1B, up 7%. GAAP operating income at the midpoints is about $79.2B, which is 4.9% below 2025. The carve-out matters: on a reported basis at guide midpoints, 2026 would be Meta's first year of operating income decline since 2022.

Street at: roughly $73.6–73.7B for Q4 revenue before the print, so the guide midpoint is about 1.2% above. Full-year 2026 and 2027 EPS consensus sit near $30.4 and $33.8.

Guidance style: Conservative on revenue, with Q3 landing above the top of its range. The expense guide was raised by exactly the charge, so the underlying cost plan did not move. We treat the 33% Q4 margin implied at midpoints as the planning case rather than a sandbag.

Analyst Q&A Highlights

What Limits Muse's User Growth

The opening question asked management to name the binding constraint on Muse adoption, whether capability, model efficiency or capacity. The answer was candid: the product has not launched outside the U.S. and Canada largely because it costs too much to serve at global scale.

Q: "But as you analyze the user behavior and the product, can you talk with us a little bit about any of the constraints to the slope of forward user growth that you are most focused on addressing, whether it is capabilities, model efficiency, capacity, etcetera, just so we understand what are the limiting factors to forward user growth that you wanna address."
— Brian Nowak, Morgan Stanley

A: "But I think 1 of the biggest constraints so far is that because this is a compute intensive service to run, we have not launched it globally yet."
— Mark Zuckerberg, CEO

Assessment: The 10 million figure is impressive precisely because it is capped. It also means the global rollout and the compute bill arrive together. Every incremental market adds serving cost before it adds revenue, which is why the 10x efficiency target is the most important operating commitment on the call.

Why Create a Non-Voting Share Now

A question tied the Class C announcement directly to capital needs, asking how management thinks about sources of capital and capital structure for the AI build. The response confirmed the funding purpose and declined to commit to any use.

Q: "And the second 1 would be with the announcement of the new class of shares tonight with earnings, any updated thoughts broadly about how you think about sources of capital and optimizing your capital structure to make sure you invest the proper amounts to capitalize on the opportunities that AI presenting in the next couple of years."
— Eric Sheridan, Goldman Sachs

A: "We do not have any specific financing plans to announce at this time. But we thought it was prudent to establish the appropriate capital structure framework to ensure that we have access to a diverse pool of funding sources."
— Susan Li, CFO

Assessment: The phrase "at this time" is the standard formulation before a financing, not a denial of one. Companies do not create a listed share class, negotiate a founder agreement and recruit a special committee for a structure they do not expect to use. We assume some Class C issuance within two years, most likely through employee compensation first.

Any Framework for 2027 Spending

A question reminded management of its past framings for the following year, such as a larger dollar increase in spending or growing operating income annually, and asked whether any applied to 2027. None was offered.

Q: "A year ago, you talked about kind of notably larger dollar increase in 2026 than 2025, and then you have also used kind of growing operating income on an annual basis as well. So just curious if there is any of those that you can kind of apply or to help us think about 2027."
— Doug Anmuth, JPMorgan

A: "But it is very dynamic, and it is still pretty early, and so we do not have anything more specific or quantitative to share until we get to the Q4 call, which is typically when we share guidance for the next year."
— Susan Li, CFO

Assessment: The refusal to restate "growing operating income" as a 2027 principle is the most informative silence on the call. Management used that framing for 2026 and could have repeated it at no cost if it intended to keep it. Combined with the 2027 commitments in the 10-Q, we read this as a warning that 2027 operating income growth is not assured.

Where 2027 Revenue Growth Comes From

A question asked whether the "strong revenue growth" framing for 2027 relied on advertising or included new AI revenue. The answer placed the new businesses as contributors next year but emphasized their scale over time.

Q: "And then for Susan, you mentioned strong revenue growth next year. Just wondering if that is advertising, or does that incorporate, some incremental AI revenues hitting next year?"
— Justin Post, Bank of America

A: "We are also, you know, building several new AI businesses, and we expect that those can begin contributing incremental revenue in 2027, but more importantly, can scale meaningfully over time, and those will include monetizing you know, our AI products through subscriptions, transactions, advertising, developing enterprise offerings as we talked about in 1 of the earlier questions."
— Susan Li, CFO

Assessment: The phrase "begin contributing" is modest language and should be modelled that way. The 2027 revenue case is the core advertising business, which management described as having a healthy ranking and recommendations pipeline. New AI revenue is a 2028 story for valuation purposes.

Muse's Cost per User

A question asked directly for metrics on Muse's per-user cost structure and how the cost curve comes down. The answer declined to give numbers and described the efficiency levers in general terms.

Q: "Mark, you touched on this earlier, but could you elaborate a little bit on maybe some high level metrics around the per user cost structure of Muse today?"
— Ross Sandler, Barclays

A: "You know, we are not we do not have sort of specific numbers to share around the cost structure of Muse today, but, of course, it is, as Mark said, pretty compute expensive."
— Susan Li, CFO

Assessment: This is the number the bull and bear cases both need and neither has. Without a serving cost per user, investors cannot translate the 10x efficiency target into a breakeven revenue per user, and so cannot judge how much of the 2027 capex Muse absorbs. Expect this to be the most-asked question on the Q4 call.

Whether Muse Monetizes Faster Than Meta's Usual Playbook

A question asked whether the cost of Muse forces monetization ahead of Meta's traditional sequence of engagement first and revenue later, and whether the commerce statistic included travel. The CEO confirmed both.

Q: "Do you think it is sort of a similar playbook here? Or it has. Does it have to be a little bit faster with Muse just given that the costs associated with it are materially greater arguably than launching those surfaces in the past."
— Mark Mahaney, Evercore ISI

A: "My guess is, I think that we are going to work on monetization somewhat sooner on this than our normal playbook. For a couple of reasons. 1 is just the size of the numbers that we are talking about in terms of both the cost and the and the opportunity, I think, make it kind of important to make progress on that in order to be able to scale it."
— Mark Zuckerberg, CEO

Assessment: This is the most investor-friendly sentence of the call, because it concedes that the compute bill imposes a revenue clock. It also means Muse's commercial terms with merchants will be tested in 2027 rather than after years of free growth, which raises both the near-term revenue potential and the risk of an early setback.

Agentic Commerce and Platforms That Resist Agents

A recurring line of questioning asked how Meta will handle aggregators and merchants that do not want agents acting on their platforms, and how agent-to-agent transactions evolve. Management described a shift toward permissioned, structured interactions and an industry protocol.

Q: "When you think about the friction points around Agentic commerce, how do you guys as a company think about solving for friction that is rising where there are aggregators or suppliers that do not want agents to interact with their platforms or their services, and how you think about lowering that friction or creating the right balance between the existing services that consumers access today relative to how this might all be changed or altered by Agentic behavior over the medium to long term."
— Multiple analysts incl. Eric Sheridan (Goldman Sachs), Doug Anmuth (JPMorgan)

A: "And I think over the long run, this will move towards more permissioned structured interactions, and that will include authenticated APIs, approved agent partnerships, or direct communications between a consumer's personal agent and a business's own agent, and we are helping establish the standards needed to support that model."
— Susan Li, CFO

Assessment: Permissioned access means Meta's commerce economics will be set in partner negotiations, not by Meta alone. The CFO acknowledged businesses that are "understandably focused on protecting customer relationships." The largest of those are the platforms with the most bargaining power. The take rate investors are waiting for will be lower than the shopping statistics imply.

Whether Muse Is Now the Center of the Company

A question asked whether the company is reorganizing around Muse after Connect and how business mandates have changed. The answer restated the personal-superintelligence vision without addressing mandates or metrics.

Q: "Mark, coming out of Connect, you know, it kind of felt like we were ushering in the next era of the company, you know, in the business reorienting around kind of Muse at the center of everything. You know, is that a fair assessment of kind of how things are going internally?"
— Mark Shmulik, Bernstein

A: "And I think as long as we continue to do that, then know, over time, the business takes care of itself as long as, you know, we continue executing well, which I am really proud of the team for how we are how we are doing on that."
— Mark Zuckerberg, CEO

Assessment: Saying "the business takes care of itself" is a founder's answer, and on the evidence of 26% revenue growth it has been a good one. Read next to the Class C structure, it is also a reminder that capital allocation at Meta answers to one holder's time horizon, and the new share class took effect by written consent without a vote of outside holders.

What They're NOT Saying

1. The $25B New Mexico judgment. Issued on the day of the print and disclosed in the next morning's 10-Q with an undisclosed accrual. It was not mentioned once on the call and no analyst asked.
2. A $105B one-quarter jump in non-cancelable commitments. Total commitments reached $454.5B with $121.4B due in 2027, yet management again declined to give 2027 capex. The filing answered the question the call avoided.
3. Muse unit economics. No serving cost per user, no purchase frequency, no take rate and no subscription conversion. The single most requested number on the call was declined.
4. Whether equity issuance is coming. Management does not have "any specific financing plans to announce at this time" for a share class whose first stated purpose is capital raising.
5. The buyback. Zero repurchases for a third straight quarter, $26.2B fewer than in the first nine months of 2025. The only buyback language in any document this quarter is the Founder Alignment Agreement's clause requiring Class C repurchases to match any Class A repurchases.
6. The $10.8B of escrowed cash. Money market funds pledged under infrastructure purchase agreements and restricted until 2028–2030 appear only in the 10-Q notes.
7. Selling compute. Last quarter's headline claim of premium offers for Meta's capacity was not repeated. Enterprise is now framed entirely as selling intelligence.
8. MTIA and custom silicon. Not mentioned for a second consecutive quarter, in the quarter gross margin fell 273bps.
9. Any return threshold on the AI build. A third consecutive quarter without one, now with a product shipped that could anchor it.
10. The engagement cost of the under-18 product changes. The 10-Q warns of "disproportionate adverse user engagement impacts" if peers do not adopt similar measures. The call did not discuss it.
11. The implied Q4 margin. Guide midpoints imply 33% against 41% a year ago, the steepest compression of the year. Management offered no margin framing.
12. The data-center venture. A new $1,037M "Proceeds from data center venture distribution" line appeared in investing cash flow with no explanation, and the El Paso venture announced in Q2 went unmentioned.

Market Reaction

• Pre-print setup: META closed at $697.52 on October 21, up 5.7% year to date against the S&P 500's +13.3%, down 4.9% over twelve months and down 5.9% over thirty days. The stock had fallen 8.5% in the six sessions before the print from $762.50 on October 13, inside a 52-week closing range of $525.72 to $777.59. The close was 25.3% above the $556.71 level at our Q2 downgrade.
• After-hours October 21: up roughly 3% to 4% at the highs on the revenue beat and Q4 guide, fading to about +1% by late evening.
• Pre-market October 22: reversed to roughly -3%.
• Reaction session, in progress: opened at $676.35, a 3.0% gap down, and traded at $653.20 by mid-morning, down 6.4%, with a $648.26 low and 14.6M shares already traded. The S&P 500 was down 1.1%, so the relative move was about five points.

The after-hours read versus the morning read. The first hour priced the press release headline: revenue above the top of guidance and a Q4 guide above consensus, which is the combination that has historically moved this stock up. The overnight reassessment priced the rest of the package. Free cash flow below zero, net cash gone, Q4 capex implied above $49B and a new share class built for raising capital are each manageable. Together, arriving on the same day as a $25B judgment, they reframed the print from a revenue-resilience story into a story about a business that may now need outside capital to fund its plans. A weak tape for megacap technology added to the move but does not explain most of it.

Positioning had already de-risked. The stock came into the print 8.5% below its mid-October level and 10% below its 52-week high, so this was not a crowded long being unwound. That makes the reaction more informative, not less: investors bid the stock up about 4% in the evening and sold it 3% below the prior close at the next morning's open, once the capital structure news had been read. At $653 the stock trades at about 19.3x the $33.8 consensus for 2027 EPS despite a materially better revenue outlook. The market is no longer paying for revenue growth that does not reach free cash flow.

Street Perspective

Debate: Is Class C a Split, a Funding Tool or a Governance Problem?

Bull view: The recapitalization is economically neutral, effectively a two-for-one split that broadens the shareholder base. A non-voting currency gives Meta flexibility for acquisitions and compensation, and the Founder Alignment Agreement ties the founder's liquidity to staying in the job and to extraordinary price targets.

Bear view: Some on the Street read the structure as the precursor to an equity raise, and expect additional capital including equity in the near term. A controlling holder with 12% of the economics can now approve large-scale non-voting issuance without diluting control, and the price-doubling dividends keep expanding the pool of non-voting stock the founder can sell without giving up votes.

Our take: The bears have the stronger argument on intent and the bulls on near-term economics. The split itself changes nothing. The option it creates matters because it arrived in the quarter free cash flow went negative, and we would rather assume dilution of 1% to 2% a year from 2027 than be surprised by it.

Debate: Does Muse Justify the Compute Bill?

Bull view: Muse is the first consumer AI agent with Meta-scale distribution, and its early retention outpaces any prior Meta launch. A merchant-paid transaction model on top of 3.64 billion daily users is a new revenue line the size of a large company, and a 10x efficiency curve turns it profitable.

Bear view: Muse is capped to two countries because it is too expensive to serve, management will not share its unit economics, and the commerce opportunity depends on partners who are actively resisting agents. The product could succeed with users and still consume more capital than it earns for years.

Our take: Muse changes the probability that the AI spend earns a return, which is why we raised fair value. It does not change the timing, because 2027 revenue contribution is described as a beginning. We give credit for product-market fit and none yet for unit economics.

Debate: Has Revenue Growth Outrun the Cost Curve?

Bull view: Revenue growth held at 26% when the guide pointed to 22%, the Q4 guide holds the mid-20s, and clean operating margin held within 1.4 points of Q2. Reality Labs losses are now falling. With revenue compounding at this pace, margins trough in 2026 and recover in 2027.

Bear view: Clean EPS grew 2% on 26% revenue growth. The margin held only because 8,700 people left in a quarter, gross margin is now falling quickly, and the Q4 guide implies an 8-point year-over-year margin decline. $121B of 2027 commitments and a refusal to restate operating income growth as a goal point to a lower 2027 margin.

Our take: The bears are right about 2026 and the debate for 2027 is closer than in July. Our base case has 2027 clean EPS growing about 10% on 20% revenue growth, which is enough to support the current multiple but not to expand it.

Debate: Is the Legal Overhang Now Quantified?

Bull view: The 52-state settlement caps the largest single exposure at about $18.6B, payable over ten years, which is less than a quarter of one year's operating cash flow. The New Mexico penalty is extreme relative to the case and likely to be reduced on appeal.

Bear view: Three jurisdictions sat outside the settlement and one has already produced a $25B judgment. Personal-injury bellwethers continue, the settlement's product changes may cost engagement, and $5.3B of contingent payments remain unaccrued.

Our take: The settlement turned an unknown into a schedule, which is a net positive. The New Mexico judgment shows that outlier outcomes are possible. We deduct about $10 per share from fair value for unaccrued exposure and treat further charges as the base case.

Model Update Needed

Valuation impact: We raise fair value to $640–700 from $610. The increase reflects higher 2027 earnings, about $33 per share against our July view of roughly flat on 2026, valued at 20–21x, less about $10 per share for unaccrued legal exposure. The 2027 estimate already absorbs the 1% to 2% annual dilution we now assume from Class C issuance. Our July target was too low: it assumed a sharper revenue deceleration than the one that happened. The midpoint of $670 sits 3.9% below the $697.52 pre-print close and 2.6% above the $653.20 mid-morning level, a market-like return consistent with Hold. All figures are on the current share count. After the Class C dividend is paid in January, per-share values halve. We would turn constructive on a disclosed Muse serving cost that supports breakeven within two years, or on a 2027 framework that restores operating income growth. We would turn negative on a Class C issuance to fund capex, or on 2027 capex above $200B without an accompanying revenue commitment.

Thesis Scorecard Post-Earnings

Overall: Thesis mixed, with the two halves moving further apart. The revenue and product half strengthened: growth held in the mid-20s instead of falling to the low 20s, Muse shipped with retention that justifies the training spend, and Reality Labs finally has a declining loss path. The profitability and capital half weakened again: gross margin broke below 80%, free cash flow went negative, net cash disappeared, commitments jumped by more than $100B, and the company created a share class whose purpose is to raise capital without diluting control.

Action: Maintain Hold, raise fair value to $640–700. Our July downgrade was right about margins and cash and wrong about revenue, and the stock rallied 25% on the part we got wrong before giving back about a third of it this morning. The revenue correction justifies a higher fair value. It does not justify an upgrade, because the case for owning Meta at 20x earnings rests on that revenue reaching shareholders, and in 2026 it is reaching suppliers, litigants and a balance sheet that has moved to net debt. Hold the position, do not add on this morning's weakness, and revisit when Meta gives a 2027 framework on the Q4 call.