Full Report

The numbers behind Meta Platforms, Inc.: as-reported financial statements and company metrics for FY2021–FY2025, traced to the source filings, opened with the share-price history those statements have to justify. Every linked figure opens the exact page of the filing it was printed on, with the statement row highlighted. Amounts in US$ millions unless noted.

Reading notes: All annual figures are 'as reported' in each fiscal year's own Form 10-K (consolidated statements of income, balance sheets, cash flows) and the Note 2 revenue / segment notes. Display unit: US$ millions (per-share and DAP excepted). Revenue breakdown uses Meta's reported disaggregation by source and segment: Advertising + Other revenue = Family of Apps, plus Reality Labs. FY2021 is cited to the MD A revenue-by-segment table (p.67) because the FY2021 10-K Note 2 shows only two comparative years; the values equal the FY2022 10-K Note 2 comparatives. Segment income (loss) from operations is from the Segment and Geographical Information note. Beginning with the FY2024 10-K (ASU 2023-07, applied retrospectively) the note expands each segment into revenue, employee compensation, and other costs; earlier filings show only revenue and income (loss) from operations. FY2016–FY2020 long-term-record figures are from the standardized SEC XBRL data feed (data/financials) and are shown without page links. FY2019–FY2020 also appear as comparative columns inside the FY2021 10-K.

Share Price — Full Available History — 14 Years

The stock closed at $669.21 on Jul 10, 2026 — up 1,650% over the window shown (+22.4% a year), trading between $17.73 and $790.00. At that close the stock trades at 28× FY2025 diluted EPS as reported below.

Source: market price feed, monthly closes, sampled from 3,555 source observations, May 2012–Jul 2026. Price return only, excludes dividends.

FY2025 at a Glance

Source: FY2025 consolidated statements [1] [2] [3] [4]. Click any linked figure to open the filing page with the row highlighted.

Revenue by Segment and Source

Revenue by Segment and Source	FY2021	FY2022	FY2023	FY2024	FY2025
Advertising	114,934	113,642	131,948	160,633	196,175
Other revenue	721	808	1,058	1,722	2,584
Family of Apps	115,655	114,450	133,006	162,355	198,759
Reality Labs	2,274	2,159	1,896	2,146	2,207
Total revenue	117,929	116,609	134,902	164,501	200,966
Total revenue growth, derived	—	-1.1%	+15.7%	+21.9%	+22.2%

Source: Note 2 — Revenue disaggregation (FY2021 from MD&A revenue-by-segment table) [5] [6] [7] [8]. Click any linked figure to open the filing page with the row highlighted.

Segment Income (Loss) from Operations

Segment Income (Loss) from Operations	FY2021	FY2022	FY2023	FY2024	FY2025
Family of Apps	56,946	42,661	62,871	87,109	102,469
Reality Labs	(10,193)	(13,717)	(16,120)	(17,729)	(19,193)
Total income from operations	46,753	28,944	46,751	69,380	83,276

Source: Segment and Geographical Information note — income (loss) from operations by segment [9] [10] [11] [12]. Click any linked figure to open the filing page with the row highlighted.

Income Statement

Source: Consolidated Statements of Income [1] [2] [3] [4]. Click any linked figure to open the filing page with the row highlighted.

Columns marked E are consensus analyst estimates shown alongside reported results for direct comparison; they are not company guidance.

Estimate source: Yahoo Finance analyst consensus, as of 2026-07-11. Estimate figures link to the consensus source, not to filing pages.

Balance Sheet

Source: Consolidated Balance Sheets [13] [14] [15] [16]. Click any linked figure to open the filing page with the row highlighted.

Cash Flow

Source: Consolidated Statements of Cash Flows [17] [18] [19] [20]. Click any linked figure to open the filing page with the row highlighted.

Long-Term Record (FY2016–FY2025)

Fiscal year	Total revenue	Income from operations	Net income	Diluted EPS	Operating cash flow	Capital expenditures
FY2016	27,638	12,427	10,217	3.49	16,108	4,491
FY2017	40,653	20,203	15,934	5.39	24,216	6,733
FY2018	55,838	24,913	22,112	7.57	29,274	13,915
FY2019	70,697	23,986	18,485	6.43	36,314	15,102
FY2020	85,965	32,671	29,146	10.09	38,747	15,163
FY2021	117,929	46,753	39,370	13.77	57,683	18,567
FY2022	116,609	28,944	23,200	8.59	50,475	31,431
FY2023	134,902	46,751	39,098	14.87	71,113	27,266
FY2024	164,501	69,380	62,360	23.86	91,328	37,256
FY2025	200,966	83,276	60,458	23.49	115,800	69,691

Source: consolidated statements across filings; older years from the standardized feed [17] [1] [18] [2]. Click any linked figure to open the filing page with the row highlighted.

Operating KPIs

KPI	FY2021	FY2022	FY2023	FY2024	FY2025
Family daily active people (DAP), December average	2,820,000,000	2,960,000,000	3,190,000,000	3,350,000,000	3,580,000,000

Source: company-reported operating metrics [21] [22] [23] [24]. Click any linked figure to open the filing page with the row highlighted.

Analyst Consensus

Estimate source: Yahoo Finance analyst consensus, as of 2026-07-11. Estimate figures link to the consensus source, not to filing pages.

Traceability

276 of 366 figures on this page (75%) link to the filing page where they are printed — click a linked figure to open the source PDF at that page with the row highlighted. Unlinked figures come from standardized data feeds or pre-filing years.

• All annual figures are 'as reported' in each fiscal year's own Form 10-K (consolidated statements of income, balance sheets, cash flows) and the Note 2 revenue / segment notes. Display unit: US$ millions (per-share and DAP excepted).

• Revenue breakdown uses Meta's reported disaggregation by source and segment: Advertising + Other revenue = Family of Apps, plus Reality Labs. FY2021 is cited to the MD A revenue-by-segment table (p.67) because the FY2021 10-K Note 2 shows only two comparative years; the values equal the FY2022 10-K Note 2 comparatives.

• Segment income (loss) from operations is from the Segment and Geographical Information note. Beginning with the FY2024 10-K (ASU 2023-07, applied retrospectively) the note expands each segment into revenue, employee compensation, and other costs; earlier filings show only revenue and income (loss) from operations.

• FY2016–FY2020 long-term-record figures are from the standardized SEC XBRL data feed (data/financials) and are shown without page links. FY2019–FY2020 also appear as comparative columns inside the FY2021 10-K.

• Long-term capital expenditures use each fiscal year's own-10-K 'Purchases of property and equipment' for FY2021–FY2025; small differences vs. the data feed for FY2021–FY2023 reflect later reclassification of PP E proceeds (see discrepancies).

• Quarterly block: single-quarter figures from the 'Three Months Ended' columns of the Form 10-Qs. Q4 quarters are omitted because Meta files no Q4 10-Q (the full year is in the 10-K) and the Q4 8-K earnings exhibits are not in the parsed corpus. Q3 FY2025 net income of $2,709M reflects a one-time $15.9B income-tax provision (One Big Beautiful Bill Act), as printed. Quarterly cash-flow statement is omitted because 10-Q cash flows are year-to-date cumulative, not single-quarter.

• 3 figure(s) differed between the data feed and the filing; the filing value is shown (see the run's metrics/metrics_tab.json for the audit trail).

---

Engine and Bet

Meta Platforms is two businesses under one ticker: a Family of Apps that turns 3.58 billion daily users into $199 billion of mostly-advertising revenue and $102 billion of operating profit, and a Reality Labs division that has lost roughly $88 billion since 2019 building the metaverse and AI hardware. The founder holds about 61% of the vote on 13% of the equity. The stock fell 77% in 2022 and has since made new highs.

What the company is

Meta earns its money in one way. Facebook, Instagram, Messenger, and WhatsApp — grouped as the Family of Apps — reach 3.58 billion daily active people, up 7% in 2025 [1], and the company sells advertising against that attention. Advertising was $196.2 billion of 2025's $201.0 billion in revenue — about 98% of the total [2]. Everything else — the messaging franchise, the payments rails, the AI assistant — exists to feed or defend that advertising engine.

Alongside it sits Reality Labs, which builds virtual- and augmented-reality hardware (Meta Quest headsets, Ray-Ban and Oakley AI glasses) and the long-dated research behind them [3]. Meta reports the two as separate segments, and the gap between them is the defining fact of the business.

One engine, one bet

In 2025 the Family of Apps produced $198.8 billion of revenue and $102.5 billion of operating income — a 52% operating margin. Reality Labs produced $2.2 billion of revenue and a $19.2 billion operating loss. The two net to $83.3 billion of operating income at a 41% margin [4].

Source: FY2025 Annual Report (Form 10-K), Consolidated and Segment Results [5].

The loss is not new, and it is not shrinking. Reality Labs has run an operating loss every year since Meta began breaking it out: about $4.5 billion in 2019, widening each year to $19.2 billion in 2025. The cumulative operating loss from 2019 through 2025 is roughly $88 billion — more than the company's entire consolidated operating profit in 2025.

Source: reported segment results — FY2021 annual report (2019–2021) [6]; FY2023 annual report (2021–2023) [7]; FY2025 annual report (2024–2025) [8].

That is the shape of the business a new owner buys: a highly profitable advertising utility, and a discretionary, founder-directed wager on the next computing platform that the utility pays for. Whether that wager is value creation deferred or value destruction repeated is the question the market keeps re-pricing.

Scale and profitability

The consolidated numbers are large in absolute terms and unusually profitable in relative terms. Revenue reached $201.0 billion in 2025, net income $60.5 billion, and diluted EPS $23.49 [9]. At roughly $669 per share and about 2.54 billion shares outstanding, the equity is worth on the order of $1.7 trillion.

Source: FY2025 Annual Report (Form 10-K), segment results and cash-flow disclosure [10] [11].

Revenue has nearly doubled since 2021, but the path was not smooth. Operating income fell from $46.8 billion in 2021 to $28.9 billion in 2022 as Apple's privacy changes and a soft ad market squeezed the core, then recovered sharply through 2025. The reported growth of 2023–2025 owes as much to cost discipline as to the top line: revenue grew and headcount and expense growth were held back.

Source: reported financials, FY2021–FY2025 Form 10-K filings; FY2024–FY2025 figures per FY2025 10-K [12].

The fallen star, already recovered

Meta is a textbook case of a market darling that the market turned on and then embraced again. The shares peaked near $382 in September 2021, then fell to $88.91 in November 2022 — a 77% drawdown — as investors recoiled from collapsing margins and the open-ended metaverse spend. They then recovered past the old high to an all-time peak of $790 in August 2025, and trade near $669 today.

Source: reported market prices (daily close), 2021–2026.

For an investor drawn to fallen stars, the important point is that the fall has already been recovered. The margin of safety that existed at $89 — when the stock traded near 10 times a depressed profit — is not the situation on offer at $669. The company itself is far stronger; the price reflects it.

Balance sheet and survival risk

A collapse scenario is hard to construct. Meta held $81.6 billion in cash and marketable securities at year-end 2025 [13] against $58.7 billion of long-term debt, most of it raised in a November 2025 note issuance [14] — a net cash position of roughly $23 billion. Operations generated $115.8 billion of cash in 2025 [15], and free cash flow was $46.1 billion after a capital-spending program that has climbed toward $70 billion a year [16].

The debt is worth watching precisely because it is new: net cash has been shrinking as Meta borrows to fund AI infrastructure while still returning capital — $26.3 billion of buybacks and $5.3 billion of dividends in 2025 [17]. But a business earning $83 billion of operating income with a net-cash balance sheet carries essentially no near-term bankruptcy risk. The risk here is not insolvency; it is overspending.

Who controls it

Meta is founder-controlled in the strongest sense the U.S. market allows. Class B shares carry ten votes to Class A's one, and Mark Zuckerberg holds almost all of the Class B — 341.8 million shares — giving him 60.8% of total voting power while owning roughly 13% of the economic interest [18]. Meta is formally a "controlled company" under Nasdaq rules and is not required to maintain a majority-independent board [19].

For an investor who prizes founder skin in the game, this cuts both ways. Zuckerberg's incentives are enormous and long-term, and the capital allocation — buy back stock, out-invest rivals, absorb years of Reality Labs losses — is the strategy of an owner, not a caretaker. It is also unaccountable: the same control that lets him fund a decade-long bet lets him keep funding it whether or not outside shareholders agree. There is no governance mechanism to stop the Reality Labs spend, and that is by design.

The question this report examines

The through-line for everything that follows: whether Meta's Family-of-Apps advertising franchise is durable enough, and priced modestly enough, to reward a new owner despite a founder with unassailable control who is directing tens of billions of dollars a year into Reality Labs and AI infrastructure with no proven return. The engine is real and the balance sheet is safe. What remains contested is how much the advertising moat is worth over time, and how much of the bet a buyer is being asked to pre-pay at today's price.

---

Financials and Estimates

Three years of statements show the advertising engine reaccelerating — revenue rose 22% in both 2024 and 2025 to $201.0 billion, and the operating margin climbed from 35% to 41% [1]. Two things move against that headline: a one-time 2025 tax charge pushed reported net income below 2024's, and capital spending that nearly doubled turned a 27% rise in operating cash flow into a 15% fall in free cash flow [2]. Consensus has revenue past $300 billion by 2027.

Three years of the income statement

The top line did something a $135-billion-revenue company is not supposed to do: it reaccelerated. After the 2022 stumble, revenue grew 21.9% in 2024 and 22.2% in 2025, reaching $200.97 billion [3]. Operating income grew faster still — from $46.8 billion in 2023 to $83.3 billion in 2025 — because costs grew more slowly than revenue after the 2023 "year of efficiency" reset the cost base [4].

Source: FY2025 Annual Report (Form 10-K), MD and A Results of Operations, consolidated statements of income data [5].

Research and development is where the money goes and where the two businesses collide. That spending reached $57.4 billion in 2025 — 29% of revenue, up from $38.5 billion in 2023 [6]. That line carries both the ad-ranking systems that drive the profit and the Reality Labs losses that drain it; the consolidated 41% margin is the net of a Family of Apps segment earning $102.5 billion of operating income against Reality Labs' $19.2 billion loss [7]. The engine and the bet, visible in one expense line.

Why 2025 earnings fell while the business grew

Reported net income is the one headline that went backwards — $60.46 billion in 2025 against $62.36 billion in 2024 — and it is the number most likely to mislead a reader skimming the financials [8]. Pretax income actually rose 21.6%, to $85.93 billion. What fell was the after-tax figure, because the effective tax rate jumped from 12% to 30% [9].

Source: FY2025 Annual Report (Form 10-K), MD and A Provision for Income Taxes [10] and Results of Operations [11].

The cause is a single, non-cash, one-time item. When the One Big Beautiful Bill Act was enacted in July 2025, Meta recorded a $15.93 billion charge in the third quarter, of which $14.03 billion was a valuation allowance written against its U.S. federal deferred tax assets [12]. Strip that discrete charge out and 2025 net income is roughly $76 billion — up about 22% on 2024 and in line with the growth in pretax profit. The law that produced the accounting charge also lowers cash taxes: management guides the 2026 effective rate to 13–16% [13].

The counter-fact worth holding: the write-down is real, and the 2024 rate of 12% was itself unusually low, flattered by prior benefits [14]. A cleaner reference point for "normal" is the 2023 rate near 18%. What would change the read is the 2026 rate landing outside the guided 13–16% band, which would signal the tax benefit is smaller than management expects.

Cash conversion meets the build

Operating cash flow tells the healthier version of the same story: $115.8 billion in 2025, up 27%, and up from $71.1 billion two years earlier [15]. Free cash flow, however, went the other way — falling to $46.1 billion from $54.1 billion — because capital expenditure nearly doubled, from $37.3 billion to $69.7 billion, on servers, data centers, and network infrastructure [16].

Source: FY2025 Annual Report (Form 10-K), Consolidated Statements of Cash Flows [17].

The build is now visible on the balance sheet as well as the cash flow statement. Property and equipment, net, rose from $121.3 billion to $176.4 billion in a single year, and total assets grew from $276.1 billion to $366.0 billion [18]. To help fund it without touching the buyback, Meta issued $30 billion of senior unsecured notes in November 2025, roughly doubling gross long-term debt to $58.7 billion [19]. Even so, $81.6 billion of cash and marketable securities against that debt leaves roughly $23 billion of net cash — the balance sheet is a source of strength, not a constraint [20].

Source: FY2025 Annual Report (Form 10-K), Consolidated Balance Sheets [21]; cash and securities per MD and A Liquidity [22].

The depreciation from this spending is beginning to land on the income statement — $18.6 billion in 2025, up from $11.2 billion in 2023 — and it will keep rising as the 2025–2026 assets are placed in service [23]. That is the mechanism by which the build converts, over time, from a cash-flow drag into a reported-margin drag.

Capital returned to owners

While it reinvests, Meta also returns cash. In 2025 it repurchased and retired 40 million Class A shares for $26.26 billion and paid $5.32 billion in dividends — about $31.6 billion returned, alongside $25.03 billion still authorized for buybacks [24]. The dividend, begun only in 2024, was raised to $0.525 per quarter in early 2025 [25].

The reduction in share count is real but modest, because stock-based compensation runs against it. Buybacks retired 40 million shares in 2025, but $20.4 billion of share-based compensation was issued the same year, so diluted shares fell only from about 2.63 billion to 2.57 billion [26]. For a value buyer, the practical reading is that a meaningful part of the buyback offsets dilution rather than shrinking the float.

What the forward estimates carry

Consensus expects the reacceleration to continue: revenue near $253 billion in 2026 and $302 billion in 2027, growth of roughly 26% and 20% (analyst consensus, as reported). Consensus EPS of $32.84 for 2026 implies a headline jump of about 40% — but most of that is the 2025 tax base normalizing; against a tax-adjusted 2025 figure near $30, the underlying step-up is closer to 10%.

Sources: FY2024–FY2025 actuals per FY2025 Annual Report (Form 10-K), Results of Operations [27]; FY2026–FY2027 figures are analyst consensus (55–59 contributing analysts), as reported.

The estimate that matters most is the one that decelerates. Consensus EPS growth drops from double digits in 2026 to about 7% in 2027 even as revenue still grows 20% — the market is already pencilling in margin compression as depreciation from the build arrives (analyst consensus, as reported). And the build is set to grow again: the company guides 2026 capital expenditure to $115–135 billion, up from $70 billion, "to support our AI efforts and core business" [28]. At that pace, free cash flow becomes a function of how fast revenue outruns the depreciation the spending creates.

Source: analyst consensus and market price as of the latest available data; P/E derived from reported and estimated diluted EPS.

On the surface, the stock trades near 28 times trailing earnings — but that multiple sits on the tax-depressed 2025 figure; on the ~$30 tax-adjusted number, or on forward estimates, it is closer to 20–22 times [29]. The mean analyst price target of about $828 sits roughly a quarter above the current $669. Whether 20-times-forward is "reasonable" for a business reinvesting at this scale is the question a dedicated valuation lens has to answer; what the financials establish is that the raw multiple overstates how expensive the earnings are, because the denominator is carrying a one-time tax hit.

---

Advertising Moat

Meta's return depends on one franchise: the Family of Apps advertising engine that supplies essentially all group profit. Its operating margin fell to 37% in 2022 when Apple's privacy change and a demand recession hit at once, then recovered to 52% by 2025 as AI-driven targeting rebuilt pricing power — the average price per ad swung from -16% to +9% over the same span [1]. That stress test, survived, is the strongest evidence the moat is real; its dependence on Meta's own AI spend is the strongest evidence against reading it as free.

How the engine earns

Advertising was $196.2 billion of Meta's $200.97 billion in 2025 revenue, and the Family of Apps produced $102.5 billion of operating income against a $19.2 billion Reality Labs loss — so the ad business is not the main story, it is effectively the whole one [2]. Meta itself decomposes that revenue into two levers: the number of ad impressions delivered, and the average price per ad [3]. Impressions track users and engagement; price tracks how well Meta can turn an impression into a result an advertiser will pay for. The two move independently, and reading them apart is how you tell an engagement story from a pricing-power story.

Source: FY2025 10-K MD&A [4]; FY2023 10-K MD&A [5]; FY2024 10-K MD&A [6].

Revenue growth is roughly the two bars combined. In 2025, impressions rose 12% and price rose 9%, and advertising revenue grew 22% [7]. Underneath the total, monetization per person keeps deepening: annual worldwide average revenue per person reached $57.03 in 2025, up 15% from 2024, across 3.58 billion daily active people [8].

The 2022 stress test

The red bars above are the whole argument. Apple's App Tracking Transparency, introduced in 2021, cut off the third-party signal Meta had used to target and measure ads; a slowing ad economy arrived alongside it. Price per ad — the direct read on how much an impression is worth — fell 16% in 2022 and a further 9% in 2023, and advertising revenue actually declined 1% in 2022, the only annual fall in the company's history as a public advertiser [9]. The Family of Apps operating margin fell from 49% in 2021 to 37% in 2022 [10].

Source: derived from segment operating income and revenue, FY2023 10-K [11] and FY2025 10-K [12].

A moat that never faces a downturn, a price war, or a technology shift is unproven. Meta's faced all three between 2021 and 2023 and the margin came back. That recovery — not the peak — is the evidence, because it shows the franchise can lose its targeting signal and rebuild the value of an impression from the inside.

What rebuilt the price

The rebuild ran on the same capability the report's central question keeps circling: AI applied to ad ranking. In early 2025 Meta deployed a new Generative Ads Recommendation Model, or GEM, an architecture it says is twice as efficient at improving ad performance for a given amount of data and compute; on Facebook Reels the model lifted ad conversions by up to 5% [13]. The commercial expression of that work is the Advantage+ suite of automated ad tools. Revenue running through Advantage+ shopping campaigns passed a $20 billion annual run-rate by the fourth quarter of 2024, growing 70% year over year [14]; by the third quarter of 2025 the run-rate through Meta's end-to-end automated solutions had reached $60 billion [15].

Sources: Q4 2024 call [16]; Q3 2025 call [17]; Q1 2025 call [18]. Run-rates in $ billions.

The same AI drives the other lever. Recommendation improvements pushed daily watch time across all video types up more than 25% year over year in the fourth quarter of 2023, feeding impression supply on surfaces like Reels [19]. The switching cost this builds is asymmetric: an advertiser can leave Meta, but not easily replicate the conversion lift of a model trained on 3.58 billion people's engagement — that scale, not the software, is the barrier a well-funded rival must clear.

Whether the margin is peaking

The recovery has a ceiling worth watching. The Family of Apps margin reached 54% in 2024 and slipped to 52% in 2025, because segment costs rose 28% while revenue rose 22% [20]. The reinvestment that defends the moat is now large enough to bend its own margin. And the price growth leans on mix: impressions are growing fastest in Asia-Pacific and on lower-monetizing surfaces such as Reels, which pull the blended price per ad down even as targeting improves it [21]. Rising average revenue per person is doing the heavy lifting that keeps the total compounding.

Tailwinds and threats

The demand side has genuine tailwinds. Impression growth is driven by users and engagement, concentrated in Asia-Pacific, and the online commerce vertical was the single largest contributor to the 2025 advertising increase — a structural shift of commerce budgets onto Meta's surfaces [22]. The broader move of advertising budgets toward AI-automated, performance-priced formats favors the platform with the most data and compute, which is the position Meta occupies. Independent sizing of the global digital-ad market was not available in this run's sources, so the tailwind is documented from the filings rather than from a third-party market forecast — a limitation worth stating plainly.

Three threats sit against it. Engagement competition is real and named in Meta's own risk factors: TikTok is cited by name as reducing some users' engagement [23]. Privacy and platform shocks like ATT can recur, and the 2022 record shows how fast they hit price. The largest legal overhang, however, moved the other way: in the FTC's monopoly suit seeking to unwind the Instagram and WhatsApp acquisitions, trial ran from April to May 2025 and on November 18, 2025 the court granted judgment in Meta's favor — though the FTC filed a notice of appeal on January 20, 2026, so the matter is not fully closed [24]. For an investor weighing a fallen star, a structural break-up risk that has been litigated and won is materially different from one still pending.

The read

On the moat playbook's test — does the advantage show in the numbers, is it company-specific, has it survived a shock — the ad franchise reads as a wide moat. A 52% segment operating margin, average revenue per person compounding 15% a year, and a price per ad that fell 25% cumulatively under ATT and then turned positive are not the marks of a business a rival can copy at will. The strongest fact against that read is that the recovery was bought: it depended on the AI and infrastructure spend that is compressing free cash flow and the segment margin itself, so the moat and the bet share a budget. What would change the read is a sustained decline in price per ad without a macro cause — the signal that AI targeting has stopped adding value — or a durable loss of engagement share to TikTok or YouTube that the impression line would show first.

---

Control and Pay

Mark Zuckerberg owns about 13.5% of Meta's economics — roughly $229 billion of stock at the current price — yet controls 60.8% of the vote. He takes a $1 salary and no equity; his reported 2025 pay of $25.1 million is almost entirely the cost of his personal security and aircraft. The economic alignment a founder-focused investor looks for is unusually large and genuine. The accountability is not: 88% of Meta's outside shareholders backed a move to one-share-one-vote in 2025, and the dual-class structure means that vote cannot pass over the founder's objection.

Ownership: the economics and the vote diverge

Meta has two share classes. Class A carries one vote; Class B carries ten [1]. Zuckerberg holds 341,823,978 Class B shares — 99.8% of the class — through a set of trusts and holding companies, plus 639,347 Class A shares held by CZ Biohub in which he has no economic interest [2]. Against 2.20 billion Class A and 342 million Class B shares outstanding, that Class B block is about 13.5% of all shares but 60.8% of total voting power [3].

Sources: 341,823,978 Class B shares valued at the $669.21 close of July 10, 2026; ownership and voting from the 2026 Proxy Statement [4]; [5].

The gap between the two bars below is the whole governance question. A holder of $229 billion of stock has his fortune tied to the same share price every other owner watches — the skin in the game is real and among the largest of any public-company founder. But the ten-vote Class B converts that 13.5% economic stake into majority control, so the alignment and the control are not the same thing and do not have to move together.

Source: 2026 Proxy Statement, Security Ownership table [6].

Outside the founder, ownership is ordinary institutional index money. BlackRock-affiliated entities hold 7.2% of Class A (2.8% of the vote) and Fidelity's FMR 6.1% (2.4% of the vote) — passive holders with no path to influence the outcome of a contested vote [7].

Pay: a $1 salary, and $25 million of security

Zuckerberg has requested a $1 annual salary and takes no bonus and no equity. In 2025 his reported total compensation was $25,125,904, essentially all of it "all other compensation" — down from $27.2 million in 2024 and up from $24.4 million in 2023 [8]. That figure is not incentive pay. The company states plainly that the substantial majority of it is the cost of personal security at his residences and during travel, an annual security allowance, personal use of private aircraft, and facility improvements to enable that travel [9]. The compensation committee declined to grant him equity in 2025 on the ground that his existing ownership already aligns him with shareholders [10].

The other named officers are paid on a conventional Silicon Valley scale, and the shape is the mirror image of the CEO's: mostly equity, little cash.

Source: 2025 Summary Compensation Table, 2026 Proxy Statement [11].

The four operating officers each received roughly $16–18 million of restricted stock on a four-year vesting schedule, placing their target pay at or above the 85th percentile of Meta's compensation peer group [12]. Olivan's larger "other" line reflects relocation and tax items tied to his role, not a second incentive [13]. The CEO-to-median-employee pay ratio is 65:1, against a median employee total of $388,200 — modest for a mega-cap only because the CEO takes almost nothing in cash or stock [14]. Shareholders approved the pay program with 89% support in 2025; the board then set the next such vote for 2028, moving say-on-pay to a triennial cadence [15].

For this reader, the pay structure is close to ideal on the alignment axis: the CEO's return comes almost entirely from the share price, not from a compensation contract. The security spending is real cash out the door, but at $25 million against $60.5 billion of net income it is immaterial to value — a rounding item, not a governance flag.

Accountability: the check that structurally cannot bind

Because Zuckerberg controls a majority of the vote, Meta is a "controlled company" under Nasdaq rules and is exempt from the requirements for a majority-independent board and an independent compensation committee. It has chosen to meet both standards anyway: 11 of its 12 directors are independent, with Robert Kimmitt as Lead Independent Director [16]. The board is credentialed — Marc Andreessen, Stripe's Patrick Collison, Exor's John Elkann, DoorDash's Tony Xu — but its authority is advisory in the situations that matter most, because it serves at the pleasure of a shareholder who cannot be outvoted.

That limit is not hypothetical. At the 2025 meeting, a shareholder proposal to recapitalize into one-share-one-vote drew 88% support from unaffiliated shareholders and still failed, defeated by the Class B votes [17]. The same proposal returns in 2026, filed by NorthStar Asset Management, which frames the imbalance directly — control of "more than 61%" of the vote on "just 14%" of the economics — and links it to a string of governance costs, including a November 2025 settlement in which directors agreed to a $190 million payment to resolve claims they failed to oversee privacy practices [18]. The board recommends a vote against, arguing the structure lets management focus on the long term [19]. Both readings can be true at once; what is not in dispute is that outside holders have no mechanism to decide between them.

The same control that funds the AI and Reality Labs build the report's central question turns on (Engine and Bet) is the control at issue here: no shareholder vote gates that spending, and the one lever holders do have — the annual meeting — has just demonstrated its limit.

What the record shows the control has done

Control is not the same as misallocation, and the capital-return record cuts the other way. In 2025 the company repurchased 40 million Class A shares for $26.26 billion and paid $5.32 billion of dividends, raising the quarterly payout to $0.525 — $31.6 billion returned in a single year, alongside the reinvestment ramp [20]. That is owner behavior, not empire-building for its own sake, and it is the strongest fact against reading the control purely as a risk.

The performance record is more mixed. On the peer benchmark in Meta's own pay-versus-performance disclosure, a $100 investment at the end of 2020 grew to $243 by the end of 2025 — a strong absolute result, and a near-quintupling from the $44 trough at the end of 2022 that a fallen-star buyer would have prized — but the same $100 in the peer group grew to $445 over the period [21].

Source: 2026 Proxy Statement, Pay Versus Performance Table (company TSR vs peer group TSR) [22].

The peer-group lag is driven by the 2022 collapse and a subsequent recovery that trailed the index off the bottom; it is the company's own chosen benchmark, and the comparison is sensitive to the 2020 start date. It is not, on its own, evidence of value destruction. It is a reminder that even a self-funded, owner-aligned founder does not guarantee peer-beating returns, and that the accountability gap is a real cost only if the capital allocation goes wrong.

The read

On the two things this chapter set out to establish, the evidence points in opposite directions and both are worth holding. Ownership and pay are as clean as a founder-alignment investor could ask: about $229 billion of personal stock, a $1 salary, no equity grants, and a security bill that is immaterial to value. Accountability is the opposite: 60.8% of the vote on 13.5% of the economics, a controlled-company structure, and an 88% outside-shareholder mandate for reform that the structure renders unenforceable. The evidence that would move the accountability read is behavioral, not structural — continued capital returns and disciplined spending would show the control being used well; a large value-destroying commitment made over clear outside objection would show its cost. The structure will not change, so the founder's judgment is the only check, and the capital-allocation record is where it will be tested.

---

AI and Reality Labs

Meta's reinvestment is not one bet but three, with three different return profiles. Reality Labs has lost about $88 billion since 2019 on revenue that has never cleared $2.3 billion a year, with no disclosed return milestone. AI infrastructure — capital expenditure of $69.7 billion in 2025, guided to $115–135 billion in 2026 — mostly serves the ad engine, where the return is measurable (Advertising Moat), but is now being locked into $237.7 billion of non-cancelable commitments. The mechanism that will test all of it, a depreciation wave, has barely begun.

Sources: FY2025 10-K, Segment Results [1]; Q4 & FY2025 earnings release [2]; Q1 2026 10-Q, Notes 6 and 8 [3] and Material Cash Requirements [4].

The report has established that the ad franchise is durable and that a founder controls it without an outside check (Control and Pay). This chapter sizes what that control is spending on, and how much of it can still be reversed if the return does not come. Meta itself flags the risk in plain terms: it expects expenses to rise as it invests in "new and unproven technologies, including AI and machine learning," and states that its AI and Reality Labs investments "have the effect of reducing our operating margin and profitability," warning that "if our investments are not successful longer-term, our business and financial performance will be harmed" [5].

Reality Labs: seven years, $88 billion, flat revenue

Reality Labs is the oldest and cleanest part of the bet, and the least productive. The segment houses Meta's virtual, augmented, and mixed-reality hardware, software, and content. Its operating loss has widened every year — from $4.5 billion in 2019 to $19.2 billion in 2025 — while segment revenue has stayed close to $2 billion the entire time, ending 2025 at $2.2 billion [6].

Sources: FY2025 10-K, Segment Results (2024–2025) [7]; FY2023 10-K, Segment Profitability (2021–2023) [8]; FY2021 10-K, Segment Profitability (2019–2020) [9].

Summed across 2019 through 2025, the operating losses total roughly $88.1 billion — $4.5B + $6.6B + $10.2B + $13.7B + $16.1B + $17.7B + $19.2B — against cumulative segment revenue of little more than $10 billion over the same span [10]. The 2025 loss alone equals 23% of consolidated operating income of $83.3 billion [11]. No filing discloses a revenue target, a break-even date, or a spending ceiling for the segment; the only stated discipline is the risk-factor caveat that the investment may not succeed [12].

This is the part of the spend that most cleanly fits the through-line's "no proven return." It is also the most reversible: Reality Labs is largely operating expense and internal development, funded out of the Family of Apps' $102.5 billion of operating income [13]. If the founder chose to cap it, the loss would compress within a year or two. The AI build, examined next, is a different matter.

The capex ramp, and a depreciation wave that has not landed

The larger number is infrastructure. Capital expenditure roughly doubled to $69.7 billion in 2025, and management guides 2026 spending to $115–135 billion, "with year-over-year growth driven by increased investment to support our Meta Superintelligence Labs efforts and core business" [14]. Unlike Reality Labs, most of this serves AI systems that already earn their keep on the ad side — the automated ad tools and ranking models detailed in Advertising Moat.

Source: derived from reported financials, FY2021–FY2025 Consolidated Statements of Cash Flows [15]; 2026 shown at the $125B midpoint of the $115–135B guidance range [16].

The gap between the two bars is the point. Capex has raced ahead of depreciation because most of the new spend sits in construction-in-progress and equipment not yet placed in service. Depreciation and amortization rose from $11.2 billion in 2023 to $18.6 billion in 2025 — real, but a fraction of the assets being built [17]. As that equipment switches on, the charge accelerates. In the first quarter of 2026, depreciation of property and equipment reached $5.68 billion, up from $3.84 billion a year earlier; the servers-and-network component alone rose 67%, to $4.38 billion [18].

Management has named this as the dominant cost pressure ahead. The CFO told investors that infrastructure would be "the single largest contributor to 2026 expense growth… driven primarily by a sharp acceleration in depreciation expense growth," and — importantly for how fast the charge builds — that Meta expects "a greater mix of our CapEx to be in shorter-lived assets than in prior years" [19]. Shorter asset lives mean the depreciation lands sooner and heavier. This is what makes AI infrastructure structurally different from Reality Labs: the spend converts into a fixed operating cost that persists whether or not the incremental revenue arrives.

The "superintelligence" framing sits on top of this. In mid-2025 the company created Meta Superintelligence Labs and Zuckerberg defined the goal as "developing superintelligence, which we define as AI that surpasses human intelligence" [20]. Beyond hardware, that ambition carries a compensation cost: the CFO cited a full year of expense for "the AI talent that we're hiring this year" as the second-largest driver of 2026 expense growth after infrastructure [21]. This is the least-proven slice: it shares the ad engine's compute, but its own return is a call option with no disclosed milestone.

From discretionary to committed

The most consequential change of the past eighteen months is not the size of the spend but its reversibility. Historically, capex was a year-by-year choice. Increasingly, it is a contract. Meta's non-cancelable contractual commitments — mostly third-party cloud capacity, servers, network infrastructure, and data centers — have climbed from $32.1 billion in September 2024 to $131.0 billion at the end of 2025 to $237.7 billion by March 2026 [22] [23] [24].

Sources: Q3 2024 10-Q, Note 12 [25]; FY2025 10-K, Note 11 [26]; Q1 2026 10-Q, Material Cash Requirements [27].

That figure understates the forward obligation. On top of the $237.7 billion, Meta disclosed $182.9 billion of operating and finance leases that had "not yet commenced" as of March 2026 — data centers, colocations, and network infrastructure with terms running as long as 30 years [28]. The pace has not slowed: in April 2026 the company added roughly $24 billion more in non-cancelable commitments, and it carries a contingent obligation to buy up to $14.7 billion of cloud capacity over five years [29]. Near-term, $42.3 billion of the commitments falls due in 2026 and $47.7 billion in 2027 [30].

Two smaller disclosures show how the build is being structured. Meta reclassified $5.0 billion of money-market funds to restricted cash tied to a multi-year purchase agreement [31], and in March 2026 it moved $1.48 billion of data-center assets — mostly construction-in-progress and land — to held-for-sale, to be contributed to a third party "for the purpose of co-developing data centers" [32]. The co-development structure keeps some of the capital off Meta's own balance sheet, but the capacity — and the obligation to pay for it — remains.

The practical consequence: a growing share of the AI build can no longer be dialed back by a change of heart. Reality Labs remains discretionary opex; a large and rising portion of the infrastructure spend is now legally committed for years. Set against Control and Pay, the accountability gap sharpens — the capital that outside holders cannot vote on is increasingly capital that management itself cannot easily unwind.

The read, the counter-case, and what would change it

The evidence points to a spend that is real, escalating, and — on the infrastructure side — hardening into multi-year contracts faster than its return is proven. Reality Labs is a self-funded, discretionary $88 billion write-off with no disclosed payback. AI infrastructure is more defensible because it powers the ad engine, but it is being committed at a scale ($237.7 billion contracted, plus $182.9 billion of pending leases) that will convert into a depreciation and cloud-cost load carried regardless of whether the superintelligence ambition pays off.

The strongest fact against a bearish read comes from management's own guidance: despite "the meaningful step up in infrastructure investment," Meta expects 2026 operating income to come in above 2025's [33]. If revenue growth stays ahead of the depreciation curve, the wave is absorbed and the build looks prescient rather than profligate. The spend is also fully self-funded — $115.8 billion of operating cash flow and a net-cash balance sheet (Financials and Estimates) — so this is a returns question, not a solvency one.

What would move the read, in either direction, is checkable in the filings:

Watch, alongside it, three lines that resolve the question before the income statement does: depreciation of property and equipment, now running at $5.68 billion a quarter and climbing [34]; the non-cancelable commitment balance, $237.7 billion and still compounding [35]; and the Family of Apps operating margin, which already slipped from 54% to 52% in 2025 as segment costs grew faster than revenue [36]. Whether the AI build is defending the moat or diluting it will show up there first.

---

Valuation

At $669, Meta trades at 28.5x reported earnings, but a one-time tax charge distorts that: the trailing multiple is 22.5x normalized and near 20x forward, the low end of the mega-cap cohort. The deep discount a value buyer would have found in late 2022 is gone; the fallen star has already re-rated. The margin of safety now sits in the durability of the earnings, not in the price.

What $669 buys

The starting arithmetic is unambiguous. Meta closed the most recent session at $669.21, and on 2.57 billion diluted shares that is a market capitalization of roughly $1.72 trillion. The company carries about $22.8 billion of net cash — $35.9 billion of cash plus $45.7 billion of marketable securities against $58.7 billion of long-term debt [1] — so enterprise value is about $1.70 trillion, barely below the equity value. Net cash is small enough, relative to size, that Meta is valued almost entirely on the earnings power of the operating business, not on its balance sheet.

Sources: current market price and share count per market data, as reported; balance-sheet cash, securities and debt from the FY2025 10-K [2].

What that $1.70 trillion of enterprise value is asked to earn against depends heavily on which line of the income statement is used — and the lines diverge more than usual this year.

The tax charge and the real multiple

Reported 2025 net income was $60.46 billion, and diluted EPS was $23.49 [3]. At $669, that is a trailing price-to-earnings multiple of 28.5x. But 2025 earnings absorbed a single non-cash charge that has nothing to do with the operating business: a $15.93 billion income-tax charge recorded in the third quarter — $14.03 billion of it a valuation allowance — triggered by the One Big Beautiful Bill Act interacting with the Corporate Alternative Minimum Tax [4]. Add it back and normalized net income is about $76.4 billion, or roughly $29.68 a share — a 22.5x multiple. Management has guided the 2026 tax rate back to 13–16% [5], so the normalized figure, not the reported one, is the fair basis for a forward comparison.

The same $669 price therefore carries six defensible multiples, and the spread between them is the whole valuation debate in one picture.

Sources: multiples derived from the FY2025 10-K income statement [6], cash-flow statement [7] and the tax-charge disclosure [8]; consensus forward EPS per market data.

On earnings-based lenses — EV/EBITDA of 16.7x, normalized 22.5x, forward 20.4x — Meta is priced like a large, profitable, moderately-growing business, not a speculation. On the cash-based lens it looks dear: at $46.1 billion of free cash flow ($115.8 billion of operating cash flow less $69.7 billion of capital expenditure [9]), the price-to-free-cash-flow multiple is 37x and the free-cash-flow yield is 2.7%. That gap between a 15x cash-flow multiple and a 37x free-cash-flow multiple is not noise. It is the capital-expenditure build (AI and Reality Labs) sitting between operating cash and free cash, and it is the single largest reason two careful analysts could reach opposite conclusions about whether the stock is cheap.

The fallen star has already re-rated

For a buyer drawn to companies the market has given up on, the relevant fact is that Meta's cheap moment has passed. The multiple compressed to a genuine trough in November 2022, when the shares changed hands near $89 against trailing earnings of $8.59 — roughly 10x, and less than 3x on an ex-Reality-Labs basis. That was the fallen-star window. Since then the stock has re-rated most of the way back: the year-end trailing multiple has climbed from 14x in 2022 to 28x now, and it currently sits above its own five-year average and only modestly below the ~34x it briefly reached at the 2025 high.

Source: derived from reported diluted EPS in the FY2021–FY2025 10-Ks [10] and year-end share prices per market data. The November-2022 intraday trough was near 10x.

The read for a value discipline is that the pessimism which made Meta a screaming buy in 2022 is no longer in the price. At 20x forward and 28x trailing, the market is not hating this company; it is paying a normal-to-full multiple for a franchise it now respects. The residual discount that remains — Meta trades cheaper than most of its mega-cap peers, at roughly 20–22x forward against Amazon near 30x and Nvidia in the mid-20s — reflects a specific worry the peers share to a lesser degree: that tens of billions of dollars of annual reinvestment may not earn its cost.

Where the multiple splits: cash today, earnings tomorrow

The forward case for the stock rests on EPS that consensus puts at $32.84 for 2026 and $34.99 for 2027, against $253 billion and $302 billion of revenue. Taken at face value, $32.84 is 40% above 2025's reported $23.49, and 20x forward earnings growing 40% looks inexpensive. But most of that headline jump is the tax charge reversing, not operating growth: measured against normalized 2025 earnings of about $29.68, consensus 2026 EPS growth is roughly 11%, even as revenue is expected to grow about 26%. The wedge between 26% revenue growth and 11% earnings growth is margin compression — the depreciation and infrastructure cost of the AI build landing in the P&L faster than the incremental revenue it is meant to produce (AI and Reality Labs).

Sources: consensus 2026 EPS and revenue per market data; normalized 2025 earnings derived from the FY2025 10-K after adding back the $15.93B tax charge [11].

This is what makes the free-cash-flow multiple, not the earnings multiple, the honest test. If the 2026 capital-expenditure guidance of $115–135 billion — up from $69.7 billion — is a temporary depression of free cash flow that reverses once the AI infrastructure earns its keep, then paying 20x normalized earnings for a business compounding revenue in the mid-20s is reasonable, and the 37x free-cash-flow multiple is a mirage. If instead the elevated capital intensity is the new steady state, then free cash flow is the truer measure of what owners receive, and 37x is the multiple that matters. Management's own guidance leans to the first reading — it expects 2026 operating income above 2025 [12] — but that is a statement about the income statement, where the depreciation lags, not about free cash flow, where the cash has already left.

The high-flyer test

Weighed against a discipline that excludes expensive stocks and demands a wide margin of safety, Meta lands in an awkward middle. It is not a high flyer in the cohort sense: at roughly 20x forward earnings and 17x EV/EBITDA, with net cash and $115.8 billion of operating cash flow, it is priced more conservatively than most of the companies it is grouped with, and nowhere near the multiples that define a speculative name. But it is not a margin-of-safety bargain either. The clean discount was a 2022 event; today's price embeds a normal multiple on earnings that are themselves flattered by a reinvestment cycle whose return is unproven. The valuation cushion is thin, and what cushion exists comes from the operating cash flow the advertising engine throws off (Advertising Moat), not from a low price.

Sources: FY2025 10-K income statement [13] and cash-flow statement [14]; 2026 guidance from the Q4 FY2025 earnings call [15].

One capital-return figure is worth holding alongside the multiple, because it bounds the downside a value buyer worries about most. In 2025 Meta returned about $31.6 billion to shareholders — $26.2 billion of buybacks and $5.3 billion of dividends [16] — while still funding the entire capital build from internal cash and holding net cash positive. A company that self-funds a $70 billion capital program, returns $30 billion, and carries near-zero bankruptcy risk is not where a value investor loses money permanently. The question the price poses is narrower: not whether Meta survives, but whether 20x normalized earnings leaves enough room if the reinvestment earns less than management expects.

---

Hyperscaler Capex

The valuation read left the case conditional on whether Meta's capital intensity is temporary. That question is answerable only against the two peers running the same build. In 2025 Alphabet and Amazon both raised capital spending at rates close to Meta's, and both saw free cash flow squeezed the same way — so the compression is an industry event, not a Meta-specific flag. Yet Meta carries the highest spend relative to its revenue base, funds it with the narrowest business, and is the only one of the three with no external cloud revenue earned on the same infrastructure.

The build is industry-wide

Meta is not spending alone. In 2025 all three of the largest U.S. AI-infrastructure builders roughly two-thirds-to-doubled their capital expenditure in a single year: Meta's purchases of property and equipment rose 87% to $69.7B [1], Alphabet's rose 74% to $91.4B [2], and Amazon's rose 59% to $131.8B [3]. The three together deployed roughly $293B of capex in one year.

Sources: Meta FY2025 10-K, Statements of Cash Flows [1]; Alphabet FY2025 10-K [2]; Amazon FY2025 10-K [3].

The cash-flow consequence was the same at each company. On a consistent basis — operating cash flow less gross property-and-equipment purchases — Meta's free cash flow fell 15% to $46.1B, Amazon's fell 77% to $7.7B, and only Alphabet's held flat at $73.3B because its operating cash flow grew fast enough (up 31% to $164.7B [2]) to absorb the step-up. The 37x price-to-free-cash-flow multiple that looked stretched for Meta on its own is, in context, the mildest compression in the group.

Sources: derived from each company's FY2025 10-K Statements of Cash Flows — Meta [1], Alphabet [2], Amazon [3].

A caveat on definitions: each company reports free cash flow slightly differently. Meta also subtracts $2.5B of finance-lease principal, giving a reported figure of $43.6B [4]; Amazon subtracts equipment sale proceeds, giving $11.2B [5]. The chart uses one consistent definition across all three so the comparison holds; the reported figures differ by rounding, not by story.

Meta carries the highest intensity on the narrowest base

Where Meta separates from the group is intensity. Capex consumed 35% of Meta's $201.0B of revenue in 2025 [6], against 23% at Alphabet (on $402.8B [7]) and 18% at Amazon (on $716.9B [8]).

Sources: derived from each company's FY2025 10-K income and cash-flow statements — Meta [6], Alphabet [7], Amazon [8].

Two things temper that reading. First, the revenue-share lens flatters Amazon: most of its $717B is low-margin retail, so capex looks light against a denominator that has little to do with the AI build. Measured against operating cash flow — the money the business actually throws off — the order inverts: Amazon reinvests 95% of its operating cash flow in capex, Meta 60%, and Alphabet 55%. Amazon is the company spending closest to its cash-generation ceiling; Meta and Alphabet retain a wider margin. Second, Meta's own guidance points the intensity higher still, not lower: 2026 capex is guided to $115-135B, roughly double 2025 and more than its entire 2025 consolidated operating income of $83.3B [16]. On the guided midpoint against 2026 consensus revenue near $253B, Meta's capex-to-revenue ratio would climb toward 50%.

Cheapest of the cohort, most concentrated bet

The value-relevant fact is that carrying the heaviest intensity has not made Meta the most expensive stock in the group — it is the least expensive. At the current price Meta trades near 28x trailing and 37x free cash flow; Alphabet trades near 33x earnings and 60x free cash flow, and Amazon near 34x earnings and several hundred times a free-cash-flow figure its own build has all but erased.

Sources: FY2025 10-Ks (revenue, capex, cash flow, cloud segment) for Meta [1], Alphabet [2] and Amazon [3]; trailing multiples derived from reported net income and current market data.

Meta also returned more of its cash to owners in proportion to what it generated. It paid out $31.6B in buybacks and dividends in 2025 [9]; Alphabet returned $55.8B [2]; Amazon paid out nothing — no dividend and no buyback in 2025 — while its free cash flow fell to near zero [3]. Two of the three fund the build and still return capital; Amazon reinvests everything.

The offsetting fact is what the spending buys, and here Meta's position is genuinely different. Alphabet and Amazon both run large third-party cloud businesses that rent the same class of compute to outside customers under contract. Amazon's AWS earned $128.7B of revenue at a $45.6B operating profit in 2025 [10], and carries roughly $244B of contracted future cloud revenue not yet recognized, with a weighted life of 4.1 years [11]. Alphabet's Google Cloud earned $58.7B at a $13.9B operating profit, growing 36% [12]. Meta has no such business. Its two reportable segments are Family of Apps and Reality Labs [13]; every server it builds is consumed internally, and the return on it must come from selling more or better-targeted ads to its own users, or from the metaverse and superintelligence ambitions that carry no disclosed payoff.

The depreciation wave is shared, and still ahead

The margin cost of these builds is largely in front of all three companies, because most of the 2025 spend is not yet running. Capex outran depreciation everywhere in 2025: Meta spent 3.7 times its property-and-equipment depreciation, Alphabet 4.3 times its $21.1B depreciation [12], and Amazon about 2.0 times its blended depreciation [3]. A capex-to-depreciation ratio above three tells you the depreciation line is set to rise for years as assets are placed in service — the same fixed cost that lands on the income statement whether or not the incremental AI revenue arrives. It is a cohort-wide phenomenon; it is heaviest, on this measure, at Alphabet and Meta.

The obligation side is harder to compare cleanly because each company defines it differently, but the direction is consistent. Meta disclosed $237.7B of non-cancelable contractual commitments at March 2026, up from $32.1B eighteen months earlier [14]. Alphabet reported $149.1B of purchase commitments and other contractual obligations at December 2025, most of it short-term [15]. Amazon states that its capex-related purchase obligations are generally cancellable [11] — its $244B backlog is contracted demand it will be paid for, the opposite of a spending lock-in. So on disclosed non-cancelable spend, Meta's commitment is both the largest and the fastest-growing, even before the leases-not-yet-commenced examined elsewhere in this report (AI and Reality Labs).

The read

Set against its two closest analogues, Meta's free-cash-flow compression is not a Meta problem — it is the shared cost of an industry-wide compute build, and Meta bears it while remaining the cheapest of the three on earnings and free cash flow and returning the most capital relative to what it makes. That is the reassuring half, and it matters for a buyer worried about paying a high-flyer price: the multiple that looked full in isolation is the most conservative in its peer group.

The unreassuring half is concentration. Meta runs the highest capex intensity of the cohort, on the narrowest revenue base, with 2026 guidance pointing higher still, and it is the only one of the three whose build earns no external, contracted cloud revenue to underwrite the spend. Alphabet and Amazon can point to paying customers renting the same GPUs; Meta's return has to be manufactured inside its own advertising funnel, plus ambitions it has not yet shown will pay. What would change this read is evidence that Meta's intensity is turning down toward the peer range while Alphabet's and Amazon's keep climbing — that would recast Meta from the heaviest spender to the disciplined one. The 2026 guidance points the other way for now.