Analysis of the Form 10-K | Fiscal year ended December 31, 2025
FY2025, closed 12/31/2025. Figures are in US$ million unless stated otherwise and come from the Form 10-K filed on January 29, 2026; governance, ownership and compensation data come from the proxy statement filed on April 16, 2026, and are attributed to it. Market data: closing price of $660.09 on December 31, 2025; 2,529,555,464 shares outstanding on January 23, 2026.
Which business generates the money, and where does it go?
Meta sells advertising placements on Facebook, Instagram, Messenger, Threads, WhatsApp and third-party apps. Marketers buy them mostly through a self-service platform, pay per impression or per action, and "do not have long-term advertising commitments with us". Revenue is recognized when the ad is shown or the action occurs, so there is almost no backlog. Reality Labs (RL) sells Meta Quest headsets, AI glasses and related content, and spends far more than it sells.
The two reportable segments are Family of Apps (FoA) and Reality Labs. Infrastructure costs are allocated by estimated usage, mostly to FoA, and there are no unallocated corporate costs.
| Revenue by segment and source | FY2025 | % | FY2024 | % | FY2023 | % | YoY 25/24 | YoY 24/23 |
|---|---|---|---|---|---|---|---|---|
| Advertising | 196,175 | 97.6% | 160,633 | 97.6% | 131,948 | 97.8% | +22.1% | +21.7% |
| Other revenue (FoA) | 2,584 | 1.3% | 1,722 | 1.0% | 1,058 | 0.8% | +50.1% | +62.8% |
| Family of Apps | 198,759 | 98.9% | 162,355 | 98.7% | 133,006 | 98.6% | +22.4% | +22.1% |
| Reality Labs | 2,207 | 1.1% | 2,146 | 1.3% | 1,896 | 1.4% | +2.8% | +13.2% |
| Total | 200,966 | 100.0% | 164,501 | 100.0% | 134,902 | 100.0% | +22.2% | +21.9% |
Geographic revenue below is allocated by customer address (Note 2). The MD&A also publishes growth by user location, a different basis that gives different numbers; the Note 2 basis is used unless stated otherwise.
| Revenue by geography | FY2025 | % | FY2024 | % | FY2023 | % | YoY 25/24 | YoY 24/23 |
|---|---|---|---|---|---|---|---|---|
| United States and Canada | 78,866 | 39.2% | 63,207 | 38.4% | 52,888 | 39.2% | +24.8% | +19.5% |
| Europe | 46,569 | 23.2% | 38,361 | 23.3% | 31,210 | 23.1% | +21.4% | +22.9% |
| Asia-Pacific | 53,817 | 26.8% | 45,009 | 27.4% | 36,154 | 26.8% | +19.6% | +24.5% |
| Rest of World | 21,714 | 10.8% | 17,924 | 10.9% | 14,650 | 10.9% | +21.1% | +22.3% |
| Total | 200,966 | 100.0% | 164,501 | 100.0% | 134,902 | 100.0% | +22.2% | +21.9% |
Meta had 78,865 employees at year-end, 6% more than a year earlier according to the MD&A. The 10-K does not report changes among its signing officers. The proxy adds that John Hegeman, Chief Revenue Officer in 2025, has since left, with his unvested RSUs canceled on his last day.
Several 2025 items have no counterpart in the comparative years: a $15,930 million tax charge from the One Big Beautiful Bill Act (OBBBA), a $13,800 million minority investment in Scale AI, and a $30,000 million bond issue. To these add an unconsolidated data center joint venture in Louisiana backed by residual value guarantees, and a longer depreciation life for servers. Each moves a different line of the 2025 accounts.
What kind of company is this, and therefore which questions decide the case?
Meta is classified here as a fast grower. Revenue grew 22.2% in FY2025 after 21.9% in FY2024. Capital expenditure absorbed 34.7% of revenue, up from 20.0% in FY2023. The operating margin, steady at 41–42% in the last two years, would fit a stalwart; the reinvestment does not, with capex at 3.74 times depreciation. The cash flow statement describes a company building capacity, not one harvesting it.
Three questions decide this case.
The first is whether the advertising engine still pays for the infrastructure being added to serve it. Ad revenue is the product of impressions and price per ad, and both are growing; but the capital base grew faster than the operating profit it produced. This question is only partly answerable: $50,521 million of the asset base is construction in progress that has not yet produced revenue, and Meta does not publish the absolute number of impressions or the price per ad, only their growth rates.
The second is how much of the next years has already been committed, and with what money. Beyond the balance sheet stand $131,046 million of non-cancelable purchase commitments, about $103,770 million of leases signed but not yet commenced, and a $45,950 million maximum exposure to the Louisiana venture. In the same year free cash flow fell 16.3% and long-term debt doubled.
The third is how much of the 2025 reported profit reflects estimates and one-off tax items rather than the business. The tax charge, the depreciation life extension, unrealized investment gains and legal accruals all moved the result. This question too is only partly answerable, because for most legal matters the document states that the reasonably possible loss "cannot be estimated".
The pair that governs the model is people and revenue per person: 3.58 billion daily active people (DAP) in December 2025 and $57.03 of annual average revenue per person (ARPP). Inside it sits the pair the company uses to explain growth, ad impressions (+12%) and average price per ad (+9%), published only as percentage changes.
Mute for Meta: inventory days (no inventory line), remaining performance obligations (contracts under one year), the PEG ratio (2025 earnings distorted by tax), and debt ratios taken alone (net debt near zero while commitments exceed equity). The deepest sections are 3-bis, 3-quinquies and 3-septies, with the year's decisions read together in 5-bis.
What does the document show about the durability and the source of growth?
The evidence the filing offers is narrow but consistent. Price per ad and impressions both rose in each year shown (3-septies), so advertisers paid more for each ad as volume increased. No customer reached 10% of revenue or receivables in any of the three years. Research and development held at 28.5% of revenue in 2025, after 26.7% in 2024 and 28.5% in 2023, a level kept in a year when net income fell.
Against this, the document itself names the limits. Marketers have no long-term commitments, "Many of our marketers spend only a relatively small portion of their overall advertising budget with us", and competitive products "such as TikTok" have "reduced some users' engagement". The DAP figure is the company's own estimate, with the limits read in 3-septies and 4-bis. Whether the advantage is widening or narrowing cannot be established from the filing.
The MD&A attributes the 9% increase in price per ad to higher advertiser demand from better ad performance, held back by impression growth in lower-monetizing regions and in Reels, which "monetizes at a lower rate than our Feed and Stories products"; the mechanism is read in 3-septies. The document provides no list prices or contract data. Currency worked against Meta: at 2024 exchange rates revenue would have been $201,380 million, a constant-currency growth of 22.4%.
Growth is organic. The 2025 acquisitions cost $4,090 million in cash and $450 million in shares, and Meta did not present pro forma results because they are "not significant". The online commerce vertical was "the largest contributor" to the advertising increase. Other FoA revenue, mostly paid WhatsApp messaging and Meta Verified, grew 50.1% to $2,584 million, still 1.3% of the total.
In 2024 the cost base grew 7.9% against revenue growth of 21.9%. In 2025 it grew 23.7%, faster than revenue. The scale effect of 2024 did not repeat. Research and development rose 30.8%, driven by an 8% increase in technical headcount, higher SBC and AI infrastructure. General and administrative rose 24.8%, mainly on higher legal-related costs, which include the lapping of a $1,550 million reduction in accrued losses that had lowered 2024. Cost of revenue grew 19.9%, a figure flattered by the depreciation life extension that cut 2025 depreciation by $2,920 million.
The FoA segment, where 82% of costs sit, shows the same reversal: its costs rose 28.0% and its operating income 17.6%. Its margin fell from 53.7% to 51.6%. Without the life extension, the consolidated cost base would have grown at least 26.8%, 4.6 points faster than revenue. With the capex already committed, depreciation is the line that will decide whether scale returns.
| FY2025 | FY2024 | FY2023 | YoY 25/24 | YoY 24/23 | |
|---|---|---|---|---|---|
| Revenue | 200,966 | 164,501 | 134,902 | +22.2% | +21.9% |
| Cost of revenue | 36,175 | 30,161 | 25,959 | +19.9% | +16.2% |
| Gross profit (computed) | 164,791 | 134,340 | 108,943 | +22.7% | +23.3% |
| Research and development | 57,372 | 43,873 | 38,483 | +30.8% | +14.0% |
| Marketing and sales | 11,991 | 11,347 | 12,301 | +5.7% | −7.8% |
| General and administrative | 12,152 | 9,740 | 11,408 | +24.8% | −14.6% |
| Income from operations | 83,276 | 69,380 | 46,751 | +20.0% | +48.4% |
| Operating margin | 41.4% | 42.2% | 34.7% | ||
| Interest and other income, net | 2,656 | 1,283 | 677 | +107.0% | +89.5% |
| Income before taxes | 85,932 | 70,663 | 47,428 | +21.6% | +49.0% |
| Provision for income taxes | 25,474 | 8,303 | 8,330 | +206.8% | −0.3% |
| Effective tax rate | 29.6% | 11.8% | 17.6% | ||
| Net income | 60,458 | 62,360 | 39,098 | −3.1% | +59.5% |
| EPS diluted ($) | 23.49 | 23.86 | 14.87 | −1.6% | +60.5% |
| Diluted shares (mn) | 2,574 | 2,614 | 2,629 | −1.5% | −0.6% |
The 3.1% fall in net income is entirely a tax effect: pre-tax income rose 21.6%. Without the $14,030 million valuation allowance recorded on enactment of OBBBA, net income would have been $74,488 million, up 19.4%. The effective rate would have been 13.3%, consistent with the company's statement that it "would have decreased by 17 percentage points to 13%".
| 12/31/2025 | 12/31/2024 | |
|---|---|---|
| Cash and cash equivalents | 35,873 | 43,889 |
| Marketable securities | 45,719 | 33,926 |
| Total liquidity | 81,592 | 77,815 |
| Long-term debt (carrying value; face 59,000) | 58,744 | 28,826 |
| Operating lease liabilities (current + non-current) | 25,153 | 20,234 |
| Finance lease liabilities | 1,184 | not available |
| Gross debt including leases | 85,081 | 49,060 (excl. finance leases) |
| Net debt including leases | 3,489 | net cash 28,755 (excl. finance leases) |
| Net cash excluding leases (liquidity − long-term debt) | 22,848 | 48,989 |
| Total stockholders' equity | 217,243 | 182,637 |
Liquidity excludes $3,227 million of restricted cash held for insurance, designated reserves and acquisition holdbacks. The balance sheet moved from a net cash position of $48,989 million (excluding leases) to $22,848 million in one year, during which the company also bought $18,330 million of private equity stakes.
| FY2025 | FY2024 | FY2023 | |
|---|---|---|---|
| Operating cash flow | 115,800 | 91,328 | 71,113 |
| Capex (purchases of property and equipment) | (69,691) | (37,256) | (27,045) |
| Principal payments on finance leases | (2,524) | (1,969) | (1,058) |
| Free cash flow (company definition) | 43,585 | 52,103 | 43,010 |
| Share-based compensation | 20,427 | 16,690 | 14,027 |
| FCF after SBC | 23,158 | 35,413 | 28,983 |
| Share repurchases (cash) | (26,248) | (30,125) | (19,774) |
| Taxes paid on net share settlement | (18,400) | (13,770) | (7,012) |
| Dividends paid | (5,324) | (5,072) | — |
| Income taxes paid in cash | 7,578 | 10,554 | 6,607 |
Free cash flow follows the company's definition, which also deducts finance lease principal, and that base is kept throughout. Operating cash flow rose $24,472 million while capex rose $32,435 million, so free cash flow fell 16.3% in a year of rising profit.
SBC was $20,427 million, 10.2% of revenue. The cash cost of settling it is visible in a separate line: Meta withheld 27 million vested shares and paid $18,400 million of employee taxes in their place, an outflow that works like a repurchase. Owner earnings are not a figure Meta publishes. Built here as net income plus depreciation and other non-cash items, minus capex and finance lease principal, they measure what the shareholder could take out of the year, with the limit that all capex, including expansion, is counted as a cost. They were $24,043 million before working capital and $23,158 million after it. The second, used in this report, equals FCF after SBC because SBC is the only item separating the two: the cash flow statement adds it back as non-cash, owner earnings treat it as a cost.
| FY2025 | FY2024 | FY2023 | |
|---|---|---|---|
| ROIC, after tax at reported rate (year-end capital) | 30.5% | 41.4% | not calculable |
| ROIC, pre-tax (year-end capital) | 43.4% | 46.9% | not calculable |
| Gross debt incl. leases / EBITDA | 0.84x | 0.58x (excl. finance leases) | not available |
| Net debt incl. leases / EBITDA | 0.03x | net cash | not available |
| Interest coverage (operating income / interest expense) | 71.5x | 97.0x | 104.8x |
| FCF / net income | 72.1% | 83.6% | 110.0% |
Invested capital is not a figure Meta publishes. It is built here as equity plus long-term debt and operating lease liabilities, minus cash, marketable securities and non-marketable investments: $192,024 million in 2025 and $147,812 million in 2024. Finance leases are left out because the 2024 balance is not disclosed, and the 2023 balance sheet is not in the document. On average capital the 2025 after-tax ROIC is 34.5%; the year-end base is kept because it is available for both years, and the four-point gap reflects capital that grew 29.9% in the year.
At the 13.3% tax rate that excludes the valuation allowance, 2025 ROIC would be 37.6%. Interest coverage uses interest net of the $535 million capitalized into construction; before capitalization it is 49.0x. Interest absorbs 1.4% of operating income, and long-term debt equals 0.97 years of net income.
At the closing price of $660.09 on December 31, 2025, the market capitalization is $1,669,734 million. The price is taken at fiscal year-end so that price and earnings share the same date; it is not updated to any later date.
| Multiple at $660.09 (12/31/2025) | Value |
|---|---|
| P/E on reported diluted EPS ($23.49) | 28.10x |
| P/E on EPS excluding the OBBBA valuation allowance ($28.94) | 22.81x |
| Price / FCF | 38.31x |
| FCF yield | 2.61% |
| Price / sales | 8.31x |
| EV / EBITDA (EV $1,673,223 million; EBITDA $101,892 million) | 16.42x |
| EV / EBIT | 20.09x |
Enterprise value adds gross debt including leases and subtracts liquidity, but not the $27,524 million of private investments. The two P/E figures bracket the tax question: what Meta reported, and what it would have reported without a non-cash charge. Price to free cash flow is higher than either because capex now absorbs most of the cash generated.
Which numbers in this balance sheet are opinions of management, and how far could they move?
The auditor, Ernst & Young, lists three Critical Audit Matters: loss contingencies, uncertain tax positions and the consolidation analysis for the Louisiana venture. They are treated first. The other notes selected are those above 5% of revenue, assets or pre-tax income or that rest on a stated estimate.
Loss contingencies (CAM, Note 11). Legal-related accruals rose from $5,523 million to $6,867 million, equal to 8.0% of pre-tax income and 1.9% of total assets. They depend on management's judgment of which losses are "probable" and "reasonably estimable"; for most matters the reasonably possible loss "cannot be estimated" but "may be material in the aggregate". What would move it is named in the note. In the Flo Health case, after a jury verdict on liability, plaintiffs seek $5,000 per class member for up to approximately 1.6 million members: $8.0 billion when the two claims are multiplied, a product of what plaintiffs assert and not an estimate of loss. Youth-related plaintiffs seek up to "the high tens of billions of dollars" in certain cases, with trials beginning in 2026, and a Spanish court awarded €542 million to news publishers, under appeal. The 2024 comparison shows how much the estimate can move: a $1,550 million reduction in accruals lowered 2024 G&A.
Uncertain tax positions (CAM, Note 14). Gross unrecognized tax benefits were $16,450 million, of which $11,230 million is carried in long-term income taxes and $11,250 million would reduce the tax provision if realized. Accrued interest and penalties add $2,600 million. The discretion lies in how much of each position is "more likely than not" to be sustained. The direction of movement is documented: the Tax Court valued the intellectual property transferred in 2010 at $7,790 million, $1,480 million more than Meta reported, which raised the 2025 provision; and in September 2025 the IRS asserted $15,890 million of additional tax for 2017–2019, "plus interest and penalties", mostly on the same transfer pricing issue. Meta states that its accrual is "adequate", and the note adds that the IRS figure does not reflect offsets such as the transition tax on accumulated foreign earnings. Even so, the amount asserted for three years is close to the whole gross reserve, which covers all years and all issues.
The Louisiana venture (CAM, Note 5). In October 2025 Meta contributed $4,300 million of held-for-sale data center assets to a venture and received a one-time distribution of $2,550 million. It kept a 20% interest, carried at $1,830 million. It has committed to fund its pro rata share of about $27,000 million of development costs, will lease the campus from 2029 (initial commitment about $12,310 million) and has given residual value guarantees with a threshold of about $28,000 million. The judgment the auditor tested is that Meta lacks "the power to direct the activities that most significantly impact the Venture's economic performance", although it provides "construction management, administrative and property management services" and will be the tenant. The maximum exposure to loss is $45,950 million, 21.2% of equity. A different primary-beneficiary conclusion would bring the venture's assets and debt onto the balance sheet; the guarantee becomes a liability only if Meta does not renew a lease and the property's value falls below the threshold.
Taxes and the valuation allowance (Note 14). OBBBA allowed immediate expensing of domestic R&D and certain capex, but the 15% corporate alternative minimum tax limits the benefit. Meta concluded that part of its federal deferred tax assets will not be used and recorded a $14,030 million valuation allowance, plus about $1,900 million for a reduced foreign-derived intangible income deduction. The total allowance rose from $3,506 million to $15,895 million, more than the $12,195 million of deferred tax assets left after it. The judgment rests on "the expected impact of future years' CAMT"; a change in the minimum tax, or in Meta's forecast of taxable income, would release part of it into earnings. The rate reconciliation shows $11,974 million for valuation allowances and the balance rose $12,389 million; the document reconciles neither with the $14,030 million the company isolates, which is the figure used here. R&D credits lowered the rate by 4.6 points and excess tax benefits on vesting RSUs by 5.0 points; the second depends on the share price at vesting and, the company warns, may not be realizable under the minimum tax. The gap between provision and cash taxes is read in 3-quater.
Depreciation life of servers (Notes 1, 6). Not a CAM, but the only change in accounting estimate the filing quantifies. From January 1, 2025, most servers and network assets are depreciated over 5.5 years. The note quantifies the effect only for assets in service at December 31, 2024: depreciation $2,920 million lower and net income $2,590 million higher, 4.3% of the year's net income. Servers placed in service during 2025 also carry the longer life, so the full effect is larger and is not disclosed. The estimate rests on management's view of how long AI and general-purpose servers remain useful; the note itself lists "technological advancements" among the reasons actual lives may differ.
Debt (Note 10). All $59,000 million of notes are fixed-rate senior unsecured, with coupons between 3.50% and 5.75% and no financial covenants. Nothing matures in 2026 and $48,750 million matures after 2030. No credit facility is disclosed.
Share-based compensation (Note 12). Unrecognized SBC was $54,810 million, to be expensed over about three years: 2.7 times the 2025 expense, already contracted. Unvested RSUs were 115.6 million shares, 4.6% of shares outstanding. The plan reserve grows automatically each year through 2035 by up to 2.5% of Class A shares; 55 million were added on January 1, 2026.
Revenue (Notes 1, 2). Advertising is reported gross because Meta controls the inventory; Payments fees are net. Deferred revenue of $1,080 million is 0.5% of revenue: there is no prepaid future the year could have consumed, and no contracted backlog.
Goodwill and intangibles (Note 8). Goodwill is $24,534 million, of which $23,028 million in FoA, after $3,796 million of additions in 2025. The annual test found no impairment, and no headroom is disclosed, so a passed test says nothing about the margin. RL's $1,506 million of goodwill sits in a segment that lost $19,193 million. Intangibles rose to $3,692 million, including $2,400 million of purchased software licenses; amortization will be $1,259 million in 2026, double the 2025 charge.
Fair value and investments (Notes 4, 5). Non-marketable equity investments grew from $6,070 million to $27,524 million, 7.5% of assets. The $13,800 million Scale AI stake and the $5,820 million Jio Platforms stake are carried at cost, adjusted only for impairment and observable transactions, so their value is an estimate until a transaction occurs. Unrealized gains on equity investments added $1,138 million to 2025 pre-tax income; without them pre-tax income would have been $84,794 million.
Leases and commitments (Notes 7, 11). Operating lease liabilities are $25,153 million, discounted at 4.3%. Leases signed but not commenced total about $103,770 million and will start between 2026 and 2030. Non-cancelable purchase commitments are $131,046 million, "mostly related to third-party cloud capacity arrangements" and servers, data centers and RL hardware, with $30,634 million due in 2026. Neither figure is on the balance sheet; together they are larger than gross property and equipment of $233,726 million.
Pensions, related parties, subsequent events. The filing describes no defined-benefit pension plan. The proxy discloses about $2,300 million paid to Broadcom, whose chief executive sits on Meta's board until the 2026 annual meeting, and $17.5 million of 2025 compensation to the Chief Revenue Officer, an immediate family member of an executive officer. Post-closing events stated in the 10-K include the FTC's appeal of the antitrust judgment and the start of the first youth-related trial in January 2026.
What sits inside the large lines?
Thirteen lines exceed 5% of total assets of $366,021 million, on one side of the balance sheet or the other. Goodwill, investments and leases are covered in 3-bis; the others are read below.
| Assets above 5% | % of assets | Liabilities and equity above 5% | % of total |
|---|---|---|---|
| Property and equipment, net | 48.2% | Long-term debt | 16.0% |
| Marketable securities | 12.5% | Accrued expenses and other current liabilities | 8.4% |
| Cash and cash equivalents | 9.8% | Operating lease liabilities, non-current | 6.3% |
| Non-marketable equity investments | 7.5% | Long-term income taxes | 5.7% |
| Goodwill | 6.7% | Retained earnings | 33.1% |
| Operating lease right-of-use assets | 5.6% | Additional paid-in capital | 26.2% |
| Accounts receivable | 5.4% |
Property and equipment is now half the balance sheet. Of the $233,726 million gross, $98,040 million is servers and network assets and $50,521 million is construction in progress, which is not depreciated and produces nothing yet. Construction in progress grew 88.5%, a net increase equal to 34.0% of the year's capex. Long-lived assets are 88.1% in the United States.
Accrued expenses of $30,729 million include compensation ($7,151 million), legal accruals, accrued capex ($4,402 million) and taxes. The largest component, "other current liabilities" of $10,387 million, grew 71.0% and the note does not break it down. The cash flow statement discloses $2,659 million of acquisitions and intangible assets still unpaid at year-end and carried in payables, accrued or other liabilities, against $172 million a year earlier; it may explain part of the increase, but the document does not say in which line it sits.
Long-term income taxes more than doubled to $21,005 million: $11,230 million of uncertain positions and $9,780 million of deferred tax liabilities. The net deferred tax position moved from an asset of $9,578 million to a liability of $9,167 million in one year, an $18,745 million swing that matches the deferred tax expense almost exactly.
Net debt in section 3 is computed on unrestricted liquidity only. Working capital in days is short and stable, on year-end balances: receivables were 35.9 days of revenue against 37.7 in 2024. Payables were 89.7 days of cost of revenue against 93.0. Payables include capex-related amounts, so the payables figure overstates days on operating costs; there is no inventory line, so the cash conversion cycle is not calculable.
Equity is $217,243 million, all attributable to Meta, with no treasury shares because repurchased shares are retired against retained earnings. Additional paid-in capital grew by exactly the year's SBC expense plus the $450 million of shares issued for acquisitions, less the $8,312 million of share withholding charged to it. Return on equity was 27.8% on year-end equity and 30.2% on average equity. Adding back the $76,051 million of repurchases charged to equity in 2023–2025 lowers the year-end ROE to 20.6%. That total is on the cost basis; cash paid was $76,147 million, and the document does not explain the gap. Repurchases before 2023 are not in the document, so the adjustment is incomplete. Retained earnings grew from $64,799 million at the end of 2022 to $121,179 million after absorbing those repurchases, the dividends and part of the share withholding. The balance is not a measure of retention by itself: in 2025 buybacks and dividends took 52.4% of net income, and 82.8% including the withholding, so the growth of retained earnings understates how much cash left the company.
About 23.0% of total assets is valued at observable prices: $79,740 million of Level 1 and Level 2 instruments and $4,391 million of bank deposits. Adding receivables brings near-cash assets to 28.4%. The remainder is historical cost that depends on useful-life estimates (property and equipment), private investments valued at cost or under the equity method (7.5%) and goodwill (6.7%).
Is the operating cash flow the business, or something passing through it?
Net income explains 52.2% of operating cash flow, and depreciation, SBC and deferred taxes together explain 49.9%. Working capital absorbed $885 million. Of the three non-cash additions, deferred taxes ($18,738 million) is the one that changed: it was a deduction of $4,738 million in 2024.
The money of others held temporarily is small. Accrued expenses and other liabilities, which include deferred revenue, added $1,514 million, or $1,500 million net of a small decrease in payables, 3.4% of free cash flow. Meta's operating cash flow does not rely on customer prepayments or on stretching suppliers.
Taxes are the real difference between the two years. The provision was $25,474 million and cash paid $7,578 million, a gap of $17,896 million that is almost entirely the non-cash valuation allowance and the deferral of federal tax through immediate expensing. Pre-tax operating cash flow, adding back taxes paid, grew 21.1%, 1.1 points below revenue; the reported operating cash flow grew 26.8% partly because cash taxes fell by $2,976 million. The company states that OBBBA "reduced our U.S. federal cash tax payments for the remainder of 2025 and future years"; the federal loss carryforwards of $16,350 million at year-end are part of the same picture and are available against future profit.
Reported capex does not cover all the capacity Meta is adding. Leases that commenced in 2025 added $7,017 million of operating and $613 million of finance right-of-use assets. Commitments for third-party cloud capacity, which will be paid as operating costs rather than capex, are inside the purchase commitments read in 3-bis. The Louisiana campus will reach Meta as a lease from 2029.
Is the capital being added earning what the existing capital earns, and what does each share own? Per-share values use diluted average shares.
| FY2025 | FY2024 | FY2023 | 25/24 | 25/23 | |
|---|---|---|---|---|---|
| Revenue per share ($) | 78.08 | 62.93 | 51.31 | +24.1% | +52.2% |
| Net income per share ($) | 23.49 | 23.86 | 14.87 | −1.6% | +58.0% |
| FCF per share ($) | 16.93 | 19.93 | 16.36 | −15.1% | +3.5% |
| Equity per share ($) | 84.40 | 69.87 | 58.26 | +20.8% | +44.9% |
The shareholder did not move backward on revenue: per-share revenue grew faster than total revenue because diluted shares fell 2.1% in two years. FCF per share is the exception. Over two years revenue per share rose 52.2% and FCF per share 3.5%, because the capex increase consumed almost all the additional operating cash.
Incremental return on capital. Between 2024 and 2025 invested capital grew by $44,212 million and operating income by $13,896 million: a pre-tax incremental return of 31.4%, against a pre-tax ROIC of 46.9% on the 2024 capital base. After tax at the reported rates, incremental NOPAT is negative ($−2,638 million), because the 2025 rate includes the valuation allowance; at the 13.3% rate excluding it, the incremental return is 24.8%. This is one pair of years, a single point and not a trend.
The figure measures timing as much as quality. Of the $44,212 million of added capital, $23,719 million is the increase in construction in progress, which by definition has no revenue yet. Excluding it, the pre-tax return on capital already in service would be 67.8%. This is our own construction, not a figure Meta publishes, and it treats all capital added outside construction as if it had worked for the full year, which the document does not allow to verify. What the document allows is the range: the capital being added earns less than the existing capital on any reading that includes construction in progress, and more on the reading that excludes it.
Reinvestment rate and implied growth. Capex minus depreciation, plus acquisitions and working capital, was $56,191 million, 95.9% of NOPAT at the reported rate. Implied growth is reinvestment divided by invested capital, 29.3%, against actual operating income growth of 20.0%. The shortfall points the same way as the incremental return: a large part of the reinvestment has not reached the income statement yet. The calculation excludes the private equity stakes seen in section 3, which would take reinvestment above NOPAT.
Read together: in 2025 Meta reinvested almost all of its after-tax operating profit and, through buybacks, withholding and dividends, returned more than its free cash flow, with debt covering the difference. This is not a rent being harvested but a reinvestment rate that only a return on new capital close to the existing one would sustain, and the document does not yet show that return.
Self-financing. Over 2023–2025, cumulative free cash flow was $138,698 million. Buybacks, share withholding and dividends took $125,725 million of it, 90.6%. Private stakes and acquisitions took another $23,472 million. Net debt issued was $48,793 million, with no repayments. Meta's growth capital came from internal cash; its capital returns and investments beyond capex were partly financed by the bond market, most of it in the last year.
Capital intensity. Cumulative capex of $133,992 million equals 82.8% of cumulative net income over three years. Capex was 2.42 times depreciation in 2023, 2.40 in 2024 and 3.74 in 2025, so the company is expanding capacity, not consuming it. Where the money for growth goes is not only capex: R&D was $57,372 million, of which $17,485 million was SBC.
If growth stopped. Meta does not separate maintenance from expansion capex. Using depreciation as an approximation of maintenance, operating cash flow minus depreciation and minus SBC is $76,757 million, 4.60% of market capitalization. This construction is ours and its limit is the depreciation figure: the life extension lowered it by $2,920 million, so the approximation is generous by at least that amount.
Operating return on enterprise value. At $660.09, EBIT divided by enterprise value is 4.98%: what a buyer of the whole company would earn on 2025 operating income before financing and taxes, with the 2026 commitments still to be paid.
Research productivity. R&D over 2023–2025 was $139,728 million, all expensed, against revenue growth of $66,064 million; capitalized software was "not material".
Do the balance sheet lines that usually move first point anywhere?
| Comparison (FY2025 vs FY2024) | Growth A | Growth B | Gap (pts) | Flag |
|---|---|---|---|---|
| Inventory vs revenue | not available | +22.2% | — | no inventory line |
| Receivables vs revenue | +16.3% | +22.2% | −5.9 | above 5 pts, favorable direction |
| Allowance vs gross receivables | not available (allowance "not material") | — | — | not calculable |
| Payables vs cost of revenue | +15.7% | +19.9% | −4.2 | none |
| Depreciation vs gross property and equipment | +20.1% | +41.9% | −21.8 | above 5 pts |
| Capitalized interest vs interest expense | +39.3% | +62.9% | −23.6 | above 5 pts, favorable direction |
| Operating cash flow vs revenue | +26.8% | +22.2% | +4.6 | none |
Three gaps exceed five points, and only one points against the company. Receivables grew more slowly than revenue, so collection improved rather than loosened, and capitalized interest grew more slowly than interest expense. Depreciation grew 21.8 points more slowly than the gross asset base: part of the gap is construction in progress, which is not depreciated, and part is the life extension. Without it, depreciation would have grown at least 39.0%, above the growth of the depreciable base (+32.7% excluding construction and land). This is the one place where an accounting estimate, not the business, separates cost from capital.
The divergence between revenue and operating cash flow, the most important comparison here, is absent. Cash flow grew faster than revenue only because cash taxes fell; on the pre-tax basis of 3-quater it grew 1.1 points less than revenue, a gap below five points.
What is the metric that governs the model, and what moves it?
Meta's model reduces to two numbers: how many people use the apps every day, and how much revenue each generates in a year. DAP counts the people who log into Facebook, Instagram, Messenger or WhatsApp on a given day, after Meta has matched multiple accounts to the same person; it is an estimate built with machine-learning models and user surveys, with a stated margin of error of about 3%. ARPP is FoA revenue in a quarter divided by the average DAP at the start and end of the quarter, summed over four quarters. Both halves come from the MD&A ("Trends in Our Family Metrics").
| FY2025 | FY2024 | Change | |
|---|---|---|---|
| DAP (average for December) | 3.58 bn | 3.35 bn | +6.9% |
| Annual ARPP | $57.03 | $49.63 | +14.9% |
| Q4 ARPP | $16.56 | $14.25 | +16.2% |
| Ad impressions delivered | — | — | +12% (2024: +11%) |
| Average price per ad | — | — | +9% (2024: +10%) |
| Advertising revenue | 196,175 | 160,633 | +22.1% |
The 2024 annual ARPP is the sum of the four quarterly values the MD&A prints. DAP growth is a December-to-December comparison, while ARPP uses quarterly averages, so the two halves are not on the same base; multiplied, they give 22.8% against FoA revenue growth of 22.4%. About two-thirds of FoA growth came from revenue per person and one-third from more people.
The second decomposition is the one Meta uses: advertising revenue equals impressions times price per ad. Impressions rose 12% and price 9%, which compounds to 22.1%, the actual growth of advertising revenue. Without published levels, the pair can only be checked in growth rates.
What moves the pair, in the document's words, is a tension between the two halves. Impressions grow fastest where they are worth least: in Asia-Pacific, driven by "increases in users and their engagement", and in Reels, which monetizes below Feed and Stories. Price rises because better targeting and measurement raise advertiser demand, and the MD&A credits "ongoing improvements to our ad performance from our ad targeting and measurement tools". The same section lists what pushes price down: signal loss from Apple's 2021 iOS changes and browser restrictions, European regulation that led to a "less personalized ads" option, and enforcement against policy-violating advertisers. In 2025 price growth slowed by one point and impression growth rose by one point, the mix the company describes when its volume growth comes from cheaper inventory.
Asia-Pacific revenue grew 19.6% by customer address and 20% by user location, the slowest region on both bases, although it is where impressions grew most: the volume added there is the cheapest. The filing does not quantify the "meaningful revenue" from resellers serving China-based advertisers, nor say in which region it is reported.
Meta does not publish: the number of impressions or the average price in dollars; revenue by app; the share of impressions or revenue from Reels, WhatsApp or Threads; time spent; the number of active advertisers; and any revenue attributed to AI features. Standard metrics that are not meaningful here are inventory turnover (no inventory), remaining performance obligations and deferred revenue as a backlog indicator (contracts under one year, $1,080 million deferred), and gross margin as a product measure, because cost of revenue mixes data center operations, partner fees and RL inventory, and Meta does not present the line.
Reality Labs gets one line: revenue of $2,207 million against a loss of $19,193 million, with no unit data published; it is a cost carried by the advertising engine, 18.2% of total costs, and the company expects the 2026 loss to be "similar to 2025".
Which risks does the document itself name, and which have already cost money? They are listed in order of severity.
Risks 1 and 4 are the ones measurable from the filing; risks 2, 3 and 6 are quantified only by what plaintiffs and regulators claim.
Were risk factors added or removed this year? The comparison is not possible: the prior-year 10-K was not provided, and the filing does not mark which risk factors are new. The only declared change in this area is methodological: in the fourth quarter of 2025 Meta updated how it estimates violating accounts, and the estimate rose to "less than 5%" of DAP.
Who controls the company, and on what terms?
Meta has two share classes: Class A with one vote and Class B with ten. According to the proxy, at April 1, 2026 Mark Zuckerberg, founder, chairman and CEO, held 341,823,978 Class B shares, 99.8% of the class and 60.8% of total voting power. He holds them through trusts and holding companies, and has pledged 12,000,000 Class B shares, about 3.5% of his holdings, "to secure certain indebtedness", under a board-approved framework. He has requested a salary of $1, receives no bonus or equity, and was paid $25,125,903 in other compensation in 2025, mostly security and private aircraft.
Other executive officers hold small positions; the CFO beneficially owned 31,429 Class A shares, of which 18,243 are RSUs releasable within 60 days, and executive equity pay is RSUs only. In the fourth quarter of 2025 the COO, the CFO and one director adopted 10b5-1 sale plans; neither document reports open-market purchases by insiders.
The proxy nominates 11 directors besides Mr. Zuckerberg; two current directors retire at the 2026 annual meeting, one of them the chief executive of Broadcom, a supplier paid about $2,300 million in 2025. Ernst & Young has audited Meta since 2007; its 2025 fees were $41,363 thousand, of which tax services were $8,379 thousand. Related-party transactions are listed in 3-bis.
What did management do with the money, and what did it produce?
Capital allocation. Over the three years shown, buybacks took $76,147 million, withholding shares for employee taxes $39,182 million and dividends $10,396 million; capex, private stakes, acquisitions and new debt are totaled in 3-quinquies. The mix changed in 2025: capex nearly doubled, buybacks fell, and private stakes went from $12 million over the two prior years to $18,330 million. The acquisitions of the comparative years left almost nothing (goodwill was unchanged in 2024); the 2025 acquisitions put 85.5% of their $4,540 million price into goodwill, justified by "advancing our AI efforts, workforce, expected synergies, and potential monetization opportunities", with no revenue or profit contribution disclosed. The cash flow line for acquisitions ($4,231 million) also includes purchases of intangible assets and is not reconciled in the document with the $4,090 million of cash consideration, and $2,659 million of the year's purchases were still unpaid at year-end.
Share repurchases. Meta repurchased and retired 40 million shares at a cost of $26,264 million, an average of $656.6 per share. That is 0.5% below the year-end close. The average uses the cost in the equity statement, which includes excise taxes; the cash outflow was $26,248 million. The share count is published in whole millions, so the average could lie anywhere between $648 and $665; the comparison with the closing price is within that margin. Dividing the average price paid by the year's FCF per share, a ratio built here and not published by Meta, gives the multiple at which Meta bought itself: 38.8 times in 2025, against 23.0 times in 2024 and 13.3 times in 2023. Spending fell from $30,125 million to $26,248 million as the multiple rose. The 2025 repurchases and dividends, $31,572 million in cash, were covered by free cash flow; adding the $18,400 million of share withholding, the total exceeded it by 14.7%.
Net dilution. In 2025 Meta issued 63 million shares to employees and removed 67 million, 27 million by withholding and the rest by repurchase. Of the $44,648 million spent on removing shares, almost all offset issuance. Over three years, 193 million shares were issued and 277 million removed, a net reduction of 84 million. That 3.2% reduction cost $115,329 million in repurchases and withholding. The buyback is in large part plugging the dilution from SBC, not returning capital.
Accounting aggressiveness. Meta publishes no adjusted earnings. Its only non-GAAP measures are constant-currency revenue and free cash flow, defined as in section 3. There are no recurring "one-off" exclusions to add up; the choices that moved 2025 are in the notes and are treated in 3-bis and 6. The impairments for facilities consolidation, $2,815 million over 2023–2024, were never excluded from any company measure.
Governance signals. Internal control is effective, with no material weaknesses, no change of auditor and no restatement. The CEO's pledge and the Broadcom relationship are disclosed and committee-approved.
Compensation and incentives (proxy). The bonus for executive officers other than the CEO is the product of base pay, a target percentage raised in 2025 from 75% to 200%, and a company performance percentage set at 115% by the compensation committee's judgment. The 2025 priorities, such as "Make our business successful", "did not have specific target levels associated with them". Equity is time-vested RSUs only. No part of executive pay is tied to a stated financial metric, to returns on capital, to per-share results or to the capital spending that defines 2025. The committee's account of 2025 cites revenue growth of 22%, DAP and the formation of "Meta Superintelligence Labs".
The year's decisions read together. In 2025 Meta issued $30,000 million of notes in November and bought no shares in the fourth quarter (Note 10; Item 5); paid $13,800 million for a minority stake in Scale AI (Note 5); placed data center assets in a venture it does not consolidate and will lease back from 2029 (Note 5); signed leases and cloud capacity contracts that sit off the balance sheet (Notes 7, 11); extended server lives (Note 1); and bought businesses and software licenses whose value went mostly to goodwill and acquired software (Note 8).
Taken together, these choices show capacity being obtained through four channels instead of one. In 2023 and 2024 it was mostly built, owned and paid from operating cash flow. In 2025 it was also financed with bonds, rented through leases and cloud contracts, and placed in a partner-funded venture outside the consolidation perimeter. What this changes in next year's accounts is readable now. Interest payments due within a year are $2,980 million against $1,165 million expensed in 2025. Lease costs will rise as signed leases commence, depreciation will rise as construction enters service, and amortization of intangibles is scheduled at $1,259 million.
Does the document admit mistakes, abandoned projects or missed objectives?
It does, in specific places. On infrastructure, the risk factors state: "We have changed, suspended, and terminated certain of these projects". On Reality Labs, they state that the investments "have generated only limited revenue and reduced our operating margin and profitability", and the MD&A records that Meta Quest sales decreased. On taxes, the notes report that the Tax Court valued the 2010 transfer "$1.48 billion higher than we reported". On the product mix, the MD&A states that Reels "monetizes at a lower rate" and will do so "for the foreseeable future".
The most direct admission sits outside the 10-K. In the proxy, the compensation committee records that "in 2025, we rebuilt the foundations of our AI program", which implies that the foundations in place before needed rebuilding; the 10-K itself contains no comparable sentence. What the filing does not contain is any statement of the return expected on the 2025 capex, or any account of the projects that were changed or terminated and what they cost.
Which accounting choices moved the result, and in which direction?
Three choices moved 2025 earnings, all discussed in 3-bis: the server life extension, the consolidation judgment on the Louisiana venture, and an accounting policy that incorporates "the expected impact of future years' CAMT" when assessing deferred tax assets. The last is a policy election, and it is what brought the expected effect of future minimum tax into the 2025 income statement as a $14,030 million charge.
ASU 2023-09 was adopted prospectively, so the 2025 tax reconciliation uses new categories and is not line-by-line comparable with 2024 and 2023.
There is no GAAP-to-adjusted gap to analyze (5-bis). The effective rate of 29.6% exceeds the 21.0% statutory rate mainly because of the valuation allowance (+13.9 points). Changes in unrecognized tax benefits added 3.6 points; RSU excess tax benefits and R&D credits, read in 3-bis, offset part of the increase. The 2026 rate the company expects, 13–16%, would return the rate to the 2023–2024 range.
Net of the three effects, the 2025 operating result is better than reported net income suggests, but it rests on a depreciation estimate applied to an asset base growing faster than revenue.
Only the first two touch the decisive questions directly: one will show what the 2025 capital produces, the other how much of the 2025 tax charge stays an exception.
What did Meta say it would do, and did it? The comparison is largely not possible. The 2024 10-K was not provided, and this filing does not restate last year's targets for capex, expenses or tax.
| Declared | Realized | Gap |
|---|---|---|
| Dividend raised 5% to $0.525 per quarter from Q1 2025 (Note 12) | $0.525 paid in each of the four quarters | none |
| Repurchase authorization of $51,280 million at 12/31/2024 (Note 12) | $26,260 million repurchased including excise taxes; $25,030 million remaining | $10 million, within the rounding of figures the note publishes to the nearest $10 million |
The 10-K sets out expectations that the FY2026 filing will allow checking: capex of $115,000–135,000 million, an effective tax rate of 13–16%, an RL operating loss "similar to 2025", and about 70% of RL operating expenses on wearables.
Suppose that three years from now this investment has produced a permanent loss of capital. What would have had to happen?
Each scenario is already visible in a line of this balance sheet; none requires an assumption the filing does not support, and none can be given a probability from it.
What would remain for the shareholder if growth stopped?
| 12/31/2025 | |
|---|---|
| Tangible equity (equity − goodwill − intangibles) | 189,017 ($74.72 per share) |
| Net working capital / current ratio | 66,886 / 2.60 (2024: 66,449 / 2.98) |
| Liquidity per share | $32.26 |
| Liquidity net of gross debt including leases, per share | −$1.38 (net of long-term debt only: +$9.03) |
| Share of assets at observable prices | 23.0% (28.4% including receivables) |
Per-share balance sheet figures use the 2,529,555,464 shares supplied. If revenue stopped growing and the operating margin returned to its three-year average of 39.4%, the liquidity would be offset by the debt and the tangible equity would be mostly servers and data centers whose value depends on their use, so the protection would come from cash generation alone, while the commitments read in 3-bis would still have to be paid.
The first question was whether the advertising engine pays for the infrastructure being added. In 2025 it paid for it in cash: operating cash flow of $115,800 million covered capex and finance lease principal of $72,215 million, and advertising revenue grew 22.1% with both price and volume up. It did not yet pay for it in returns. The capital added in the year earned 31.4% pre-tax against 46.9% on the capital already in place. Over two years FCF per share barely moved while revenue per share rose by half (3-quinquies). The document cannot settle this question because $50,521 million of the added capital is not yet in service.
The second question was how much of the future is already committed and with what money. The answer is more than the company's equity: purchase commitments and unstarted leases total $234,816 million, against equity of $217,243 million, with the venture guarantees on top. In 2025 the money came from operations and from debt, which doubled to $59,000 million at face value, while buybacks, withholding and dividends together exceeded free cash flow by 14.7%.
The third question was how much of the 2025 profit reflects estimates. Reported net income is understated by a $14,030 million tax charge and overstated by at least $2,590 million from the depreciation life extension. Unrealized investment gains added a further $1,138 million before tax. The legal exposure cannot be measured from the filing.
At $660.09 the shareholder pays 28.10 times reported earnings, 22.81 times earnings without the valuation allowance and 38.31 times the cash the business left after investing in 2025. The price therefore rests on cash from capacity that the 2025 accounts do not yet show. The reading holds if the incremental return on capital rises once construction is in service and FCF per share resumes growing; it fails if depreciation and lease costs grow faster than revenue, or if the tax and youth disputes resolve near the amounts claimed.
The FY2026 statements will allow checking six items. Depreciation against revenue, with $50,521 million of construction entering service, will show whether scale returns. The growth of price per ad (+9% in 2025) against impressions (+12%) will show whether new capacity raises what advertisers pay or only the volume. FCF per share ($16.93) will show whether the shareholder captures the growth. Unrecognized tax benefits ($16,450 million) and the 2010 Tax Court decision will size the tax risk. Legal accruals ($6,867 million) will be the first line to reflect the youth trials. Commitments, unstarted leases and the venture's consolidation status will show how much capacity is still to be paid for.
The filing does not state the return Meta expects on its capex and commitments, nor the split between maintenance and growth capex. It does not publish impressions or price per ad in levels, revenue from Reels, WhatsApp, AI features or China-based advertisers. It does not explain "other current liabilities" of $10,387 million or "other financing activities" of $2,220 million, who funds the other 80% of the venture, or which projects were "changed, suspended, and terminated". Each would change the answer to the first two questions.
This report is based on the Form 10-K of Meta Platforms, Inc. for the fiscal year ended December 31, 2025, filed on January 29, 2026, and, for governance, ownership, related parties and compensation, on the DEF 14A filed on April 16, 2026. Multiples use the closing price of $660.09 on December 31, 2025 and 2,529,555,464 shares outstanding at January 23, 2026, both supplied with the request. The report is for analytical purposes only and does not constitute investment advice.
1-bis
2
3
3-bis
3-ter
3-quater
3-quinquies
3-sexies
3-septies
5-bis
8-ter
9
The full data table behind this report — every line item with its source in the filing and the arithmetic checks — is published separately: Meta Platforms, Inc. FY2025 data table.