META — Q3 2026 Financial Report Analysis

A $9.89 billion settlement halved the margin; underneath it, ads grew 26% and costs grew 36%

Meta Platforms' third quarter of 2026 (the three months to 30 September) has two very different headlines, depending on which line you read. Revenue rose 26% to $64.60 billion, almost all of it from advertising. Income from operations, meaning what the business earns from running its apps and hardware before interest and tax, fell 37% to $12.99 billion. The reason is one item: a $9.89 billion charge for the August 2026 agreement with 52 US state attorneys general to settle litigation over teen mental health and social media, an agreement that "provides for payments of up to approximately $18.59 billion."

Net income, meanwhile, rose 264% to $9.86 billion and diluted earnings per share (EPS, profit divided by every share that could exist, including employee stock awards) to $3.83 from $1.05. That jump is not a sign of a great quarter. A year ago Meta booked a one-time, non-cash tax charge of $15.93 billion, which left only $2.71 billion of net income. Both years' headline profit figures are distorted by a single large item, in opposite directions.

This analysis is written from Meta's earnings release (Exhibit 99.1 to the 8-K filed 21 October 2026). The full quarterly report (10-Q) has not been filed yet, so details such as which segment the settlement charge sits in, the regional revenue split and management's line-by-line explanations of cost growth are not available until it is.

Meta reports two segments. Family of Apps is Facebook, Instagram, Messenger and WhatsApp, and makes essentially all of the money. Reality Labs is the virtual/augmented-reality hardware and software unit (Quest headsets, AI glasses), which loses money by design. "Daily active people" counts unique users of at least one of those apps on an average day. ARPP is our calculation, on the same basis as our Q2 report: Family of Apps revenue for the quarter divided by the average of daily active people at the start and end of the quarter (3.60bn in June and 3.64bn in September 2026; 3.48bn and 3.54bn a year earlier). The "excluding one-off" EPS figures are Meta's own: $7.40 is what it says Q3 2026 EPS would have been without the settlement charge, and $7.25 is what it said Q3 2025 EPS would have been without the tax charge.

The ad business: still growing fast, but the mix has shifted

Advertising revenue grew 25.9% to $63.06 billion, and currency was not a factor: at last year's exchange rates, revenue growth was also 26% (that is what "constant currency" means). For the first nine months, reported growth of 29% compares with 27% at constant currency, so most of this year's currency tailwind has now faded.

The way the growth is produced has changed. In Q3 2025, Meta showed 14% more ads at prices 10% higher. In Q2 2026 it was 14% more ads at prices 12% higher. This quarter it was 10% more ads at prices 14% higher. More of the growth now comes from advertisers paying more per ad, which Meta has attributed in past filings to better ad targeting and measurement. Price-led growth is generally higher quality, because it does not need more ads squeezed into people's feeds. But volume growth slowing from 14% to 10% in one quarter is a change worth following; the 10-Q should say whether it reflects fewer new ad slots (for example in Reels) or regional mix.

User growth is not the engine. Daily active people rose 3% to 3.64 billion. Revenue per person rose about 22.5%. Family of Apps "other revenue" (mainly paid business messaging on WhatsApp and subscriptions) grew 59% to $1.10 billion. It is still only 1.7% of revenue, but it is Meta's only meaningful non-advertising income that is growing.

Where the costs went

Total costs and expenses rose 68% to $51.61 billion. Without the settlement charge they were $41.72 billion, up 36%, still well ahead of 26% revenue growth:

• Research and development: $20.05 billion, up 32%. This is where most AI model work, AI engineering salaries and much of the share-based pay sit. It actually fell from $21.66 billion in Q2, which included severance for the May 2026 job cuts.
• Cost of revenue: $13.37 billion, up 45%. This is mostly the cost of running the data centres and servers that serve the apps and their ads. It is the fastest-growing cost line, which fits with depreciation (the annual cost of equipment spread over its life) rising 44% to $7.15 billion as new capacity comes into service.
• General and administrative: $14.86 billion, against $3.51 billion. Without the $9.89 billion charge it was $4.97 billion, still up 41%. The release does not explain the rest of the increase; Q2's legal charges of $2.40 billion suggest legal costs are a recurring element.
• Marketing and sales: $3.33 billion, up 17%, the only major line growing slower than revenue.

The May 2026 headcount reduction has now gone through. Headcount was 66,775 at the end of September, down 15% from a year ago and down from 75,472 at the end of June, consistent with Meta's statement in its Q2 10-Q that most of the roughly 8,000 affected employees would no longer be counted by the end of Q3. Even with about 11,700 fewer people than a year ago, R&D spending rose 32%: the cost of AI infrastructure and of the remaining staff is rising faster than the headcount savings.

Takeaway: Strip out each year's one-off and Meta's earnings per share grew 2% ($7.40 against $7.25) on 26% more revenue. The ad business is healthy, and growth is now led by higher prices. But nearly all of that extra revenue is being absorbed by AI-related costs before it reaches profit, and the cash picture is tighter still: free cash flow turned negative this quarter, and guidance implies Q4 capital spending of $49–59 billion, a sharp step up from Q3's $35 billion.

What the headline numbers hide

Both years' profits are distorted, so the clean comparison is narrower than either headline. Reported net income rose 264% and operating income fell 37%. Neither describes the underlying business. Meta's own adjusted figures say operating income would have been $22.88 billion without the settlement charge, 11.4% above last year's $20.54 billion, and EPS $7.40 against last year's tax-adjusted $7.25. These adjustments are Meta's; there are no other exclusions in them.

Taxes went the other way under the surface. Without the settlement charge, the effective tax rate (the share of pre-tax profit paid as tax) would have been 17%, against 14% for Q3 2025 without its tax charge. The reported 24% is higher still because, by Meta's own numbers, the $9.89 billion charge reduced EPS by $3.57 a share, roughly $9.2 billion of net income on 2,573 million diluted shares: about 93% of the pre-tax charge, which suggests little or none of it is tax-deductible. A higher tax rate is one reason clean EPS grew only 2% while clean operating income grew 11%.

Investment income disappeared. "Interest and other income" fell to $13 million from $1.13 billion. A year ago it included a $962 million unrealized gain on equity investments; this quarter that line was an $18 million loss. Long-term debt is $80.92 billion, up from $58.74 billion at the end of 2025 after a bond issue that raised $24.91 billion earlier this year, so interest costs are higher too. The cleaner comparison is operating income, not pre-tax income.

Buybacks did nothing for EPS. Diluted shares were 2,573 million against 2,572 million a year ago. Meta has repurchased no stock in 2026 (it spent $26.25 billion on buybacks in the first nine months of 2025), so EPS moves almost exactly with net income.

Cash flow from operations looks strong partly because the settlement has not been paid yet. Operating cash flow was $32.79 billion, 3.3 times net income. Share-based compensation ($7.36 billion, up 33%, 11.4% of revenue) and depreciation ($7.15 billion) are large non-cash costs that explain part of the gap. Another part is that the charge was booked but not yet paid: the change in "other liabilities" added $7.44 billion to operating cash flow in the quarter, and that balance-sheet line rose to $13.35 billion from $4.25 billion at the end of 2025. That is consistent with most of the settlement cash still to come, which would weigh on cash flow in later quarters. The release does not give a payment schedule, nor does it explain the gap between the $9.89 billion charge and the "up to approximately $18.59 billion" headline figure; the 10-Q should.

Capital spending is now larger than operating cash flow. Capital expenditure including finance-lease payments was $34.91 billion, against $32.79 billion of operating cash. Free cash flow for the first nine months was $11.05 billion, down from $29.51 billion, and $4.06 billion of that went to dividends. Cash and marketable securities were $82.45 billion, against $80.92 billion of long-term debt, so Meta's net cash position is now close to zero. Another $12.54 billion sits in restricted cash in "other assets" (up from $1.61 billion a year ago); in Q2 Meta said most of the restricted money was escrow for multi-year infrastructure purchase agreements.

Receivables are not a concern. Accounts receivable rose 15% since December to $22.72 billion, slower than the 29% revenue growth for the nine months.

Guidance moved on cost and capex, not revenue. Compared with what Meta said at Q2: the full-year expense outlook rose to $175–179 billion to include the settlement charge; the capex range was narrowed to $135–145 billion from $130–145 billion (the top was unchanged, the floor rose); the Q4 tax rate outlook stayed at 15–17%; and the Reality Labs loss outlook improved, from "similar to 2025" to a decline, to $18–19 billion.

Reality Labs: shrinking losses, shrinking revenue

Reality Labs lost $4.13 billion, 6.8% less than a year ago, on revenue of $443 million, down 5.7% from $470 million. Revenue grew 16% in Q2, so the unit went from growth to decline in one quarter; the release doesn't say why, and the 10-Q's split between AI glasses and Quest headsets is what will answer it. For the first nine months the loss was $12.78 billion against $13.17 billion.

The forward guidance is the bigger change. Meta now expects Reality Labs' 2026 loss to be $18–19 billion (implying a Q4 loss of $5.2–6.2 billion, larger than Q3's) and losses to "narrow next year to around $15 billion as we reallocate investment toward other priority areas, primarily AI." For the first time in several years, Meta is guiding its metaverse spending down, explicitly to fund AI.

A new class of shares, and what it changes for readers of future results

The release also announced a new non-voting Class C common stock. From January 2027 (the certificate is set to become effective on 19 January 2027), every holder of Class A or Class B stock will receive one Class C share for each share they own. Meta expects the Class A share price to "reflect the effect of a two-for-one stock split" once the dividend is paid; the record and payment dates have not been set.

Why it matters: Meta says the new class is for "capital raising, employee compensation, potential acquisitions, and other general corporate purposes," given "the scale of investment required" for AI. Shares without votes can be issued for those purposes without diluting the voting control of Mark Zuckerberg, who controls the company through Class B shares. The board also signed a "Founder Alignment Agreement" with him covering his sales of Class C stock, his continued service and conditions for future Class C dividends. The stock dividend does not change anyone's voting power today. For anyone tracking Meta's numbers: once it takes effect, the share count roughly doubles, so per-share figures such as EPS and dividends per share from 2027 will be about half of what they would otherwise be, and comparisons with earlier quarters will need adjusting.

Did last time's read hold up?

Our Q2 2026 report named Family of Apps operating margin as the number to watch: a recovery into the mid-40s would make Q2 look like a spending peak, while staying near 40% on clean costs would mean revenue growth was no longer enough to grow profits. The reported Family of Apps margin fell to 26.7% from 49.2%. The release doesn't say which segment carries the $9.89 billion charge; if it is all in Family of Apps, the clean margin would be about 42%. That is above Q2's reported 39% but short of the mid-40s, and at company level the clean operating margin slipped to 35.4% from about 37% in Q2 (our own add-back of Q2's legal charge and severance). So the cautious reading mostly held: spending did not peak in Q2. One part did not: clean operating income still grew 11% year over year, so profits are growing, just much more slowly than revenue. Our Q2 expectation that second-half free cash flow would be near zero or negative was borne out in Q3 (−$2.12 billion), and the headcount savings arrived on schedule.

Outlook

What management guided:

• Q4 2026 revenue of $73–76 billion, with currency about a 1-point headwind. Against Q4 2025 revenue of $59.89 billion (full-year 2025 revenue of $200.97 billion less the first nine months' $141.07 billion), that is growth of about 22–27%, broadly the pace of Q3.
• Full-year 2026 total expenses of $175–179 billion, including the settlement charge. With $127.08 billion spent in the first nine months, that implies Q4 expenses of about $47.9–51.9 billion. Set against the revenue guide, Q4 operating income would land between about $21 billion and $28 billion, around $24.6 billion at the midpoints, roughly flat with Q4 2025's $24.75 billion. Excluding the Q3 charge, Meta still expects 2026 operating income to exceed 2025's $83.28 billion.
• 2026 capital expenditure of $135–145 billion. With $85.82 billion spent through September, Q4 alone would be $49–59 billion.
• Q4 tax rate of 15–17%, absent changes to tax law.
• Reality Labs losses of $18–19 billion in 2026 and around $15 billion in 2027.
• 2027: no revenue outlook, but Meta expects to "sustain strong overall revenue growth throughout 2027" and to "continue investing aggressively." Management also flagged "a number of youth-related trials scheduled in the coming months in the U.S."

Our read. The ad business is not the problem. Growth held at 26% at constant currency, and higher prices now do more of the work, which is the better kind of growth. The problem is how little of it reaches shareholders. On clean numbers, EPS grew 2% on 26% more revenue, and Q4 guidance implies operating profit roughly flat year over year even before any further legal costs. Q4 capital spending of $49–59 billion against roughly $33 billion a quarter of operating cash flow means free cash flow is likely to be negative again, while settlement payments are still to come and net cash is already about zero. That is the context for the new Class C shares: they give Meta a way to raise money and pay staff in stock without touching founder control. The 2027 Reality Labs cut, about $3–4 billion less loss than 2026, helps but is small next to a capex budget that runs well over $100 billion a year.

The number to watch in Q4 2026 (results expected in late January 2027) is operating income against the roughly $21–28 billion range implied by guidance, together with Meta's first 2027 capex figure. If operating income comes in near the top of that range while 2027 capex growth slows, this year looks like the peak of the cost squeeze. If it comes in near the bottom and 2027 capex rises again, Meta is committing to several more years of revenue growing faster than profit, with the new share class as a likely way to pay for it. Litigation is the other open item: the scheduled youth-related trials, and how much of the "up to $18.59 billion" settlement still has to be expensed or paid.