Analysis of the Form 10-K | Fiscal year ended December 31, 2025
FY2025, closed on 31/12/2025. All amounts in US dollars. Every figure comes from the Form 10-K filed on February 5, 2026, unless it is attributed to the 2026 Proxy Statement, which is used only for ownership, compensation and related parties. Market data were supplied with the request and are stated where the first multiple appears.
Alphabet sells attention and computing capacity. Most of the money comes from advertisers who pay when a user clicks or views an ad next to a search result, a YouTube video or a partner's website. The rest comes from consumers paying for subscriptions, apps and devices, and from companies renting infrastructure, AI models and software through Google Cloud, billed on consumption or by subscription.
The document reports three segments. Google Services holds advertising, subscriptions, platforms and devices. Google Cloud holds Google Cloud Platform and Workspace. Other Bets groups the non-Google businesses, Waymo among them. A fourth, unallocated line, "Alphabet-level activities", carries shared research on the company's general AI models, plus corporate costs and certain fines and settlements.
| Revenue by type (US$ mn) | FY2025 | % of total | FY2024 | FY2023 | YoY FY2025 |
|---|---|---|---|---|---|
| Google Search & other | 224,532 | 55.7% | 198,084 | 175,033 | +13.4% |
| YouTube ads | 40,367 | 10.0% | 36,147 | 31,510 | +11.7% |
| Google Network | 29,792 | 7.4% | 30,359 | 31,312 | −1.9% |
| Subscriptions, platforms and devices | 48,030 | 11.9% | 40,340 | 34,688 | +19.1% |
| Google Services | 342,721 | 85.1% | 304,930 | 272,543 | +12.4% |
| Google Cloud | 58,705 | 14.6% | 43,229 | 33,088 | +35.8% |
| Other Bets | 1,537 | 0.4% | 1,648 | 1,527 | −6.7% |
| Hedging gains (losses) | (127) | 0.0% | 211 | 236 | n.m. |
| Total | 402,836 | 100.0% | 350,018 | 307,394 | +15.1% |
Advertising was 73.2% of revenue in FY2025, against 77.4% two years earlier. The shift is toward lines that the document itself says carry lower margins: Cloud, subscriptions and devices.
| Revenue by geography (US$ mn) | FY2025 | % of total | FY2024 | FY2023 | YoY FY2025 |
|---|---|---|---|---|---|
| United States | 194,229 | 48.2% | 170,447 | 146,286 | +14.0% |
| EMEA | 117,152 | 29.1% | 102,127 | 91,038 | +14.7% |
| APAC | 67,680 | 16.8% | 56,815 | 51,514 | +19.1% |
| Other Americas | 23,902 | 5.9% | 20,418 | 18,320 | +17.1% |
| Hedging gains (losses) | (127) | 0.0% | 211 | 236 | n.m. |
| Total | 402,836 | 100.0% | 350,018 | 307,394 | +15.1% |
Alphabet had 190,820 employees at December 31, 2025; the document does not give the prior-year headcount. The 10-K is signed by Sundar Pichai as CEO and Anat Ashkenazi as CFO; the Proxy Statement dates her appointment to July 2024, so no signatory changed during FY2025.
Several items appear this year that the comparative columns do not contain: credit derivatives backing third-party data center leases (zero a year earlier), euro-denominated debt, guarantees for energy projects, two pending acquisitions and a final judgment in the US Search case. Together they describe a company that has started to finance and guarantee capacity outside its own balance sheet.
Alphabet is classified as a solid company (stalwart) whose decisive questions come from the capital-intensive side of the business. Two figures fix the category: revenue grew 15.1% in FY2025 and the operating margin held at 32.0%. Mid-teens growth is not the high double-digit growth of a fast grower, the margin is established rather than forming, and no price or volume series in the document drives it as in a cyclical. What makes the case unusual is the capital: capex was 4.33 times depreciation, and commitments not yet on the balance sheet exceed the year's capex. The stalwart's questions, what is being paid for and how durable the margin is, must therefore be asked of the capital being committed.
Three questions decide the case. The first is what the capital now being committed earns, against what the capital already in place earns. The document answers this only in part. It allocates no assets or capex to segments, does not separate maintenance from expansion and publishes no capacity in physical units; it does disclose the US$78.6 billion of plant not yet in service, which allows a second reading on capital that is producing.
The second is how much of FY2025 profit is operating profit. Net income rose 32.0% and operating income 14.8%; the gap is a US$24.1 billion gain on equity securities, mostly unrealized revaluations of private holdings. The document answers this fully, and reports a further US$32.0 billion of such gains in January 2026.
The third is whether Search, 55.7% of revenue, still grows by volume and price together. The document publishes only percentage changes for clicks and price, no query volumes, no revenue from AI features and no cost estimate for the court-ordered remedies. This question is also only partly answerable, and is treated as decisive anyway.
The engine of the model is described by two pairs. For Search, the pair is paid clicks and cost-per-click, with the TAC rate as the share of that revenue paid away to partners. For Cloud, the document publishes no units, so the pair becomes Cloud revenue and the Cloud operating margin, with the US$242.8 billion revenue backlog as the balance that governs it.
Three quantities are mute for this company and are not used to judge it: inventory days (inventory is not disclosed), net debt to EBITDA (the company holds net cash), and the dividend (13.7% of free cash flow). A price-to-earnings-to-growth ratio is omitted, because earnings growth inflated by revaluation gains would measure the gains rather than the business.
The report goes deepest in 3-quinquies (returns and reinvestment), 3-bis together with 6 (the revaluation gains), 3-septies (the two pairs) and 5-bis (the capital decisions of the year).
The document shows little about the advantage, and only that is used. No customer exceeded 10% of revenue in any year. Search & other grew 13.4% to US$224.5 billion, with paid clicks and cost-per-click both rising. Research spending rose 23.8% to US$61.1 billion, 15.2% of revenue.
The company calls its "AI-optimized infrastructure" "a key differentiator", a claim the document does not measure, and it gives no data showing whether the advantage is widening or narrowing. It does record that a US court found against Google on Search distribution in August 2024, and that the December 2025 remedy requires it "to share certain search data with and offer syndication services to certain competitors". Part of what has protected Search is now under a court order whose cost is not estimated.
Cost-per-click on Search rose 7% in FY2025 while paid clicks rose 6%, so price contributed slightly more than volume to the growth of the main line. On Google Network, cost-per-impression also rose 7%, but impressions fell 7% and revenue declined 1.9%. The document attributes price changes to "device mix, geographic mix, advertiser spending, ongoing product and policy changes, product mix, property mix, and changes in foreign currency exchange rates", so the 7% cannot be read as a price increase imposed on the same product.
The TAC rate, the share of advertising revenue paid to partners, fell from 20.7% to 20.3%. The company attributes this to mix, while rates on each channel were "substantially consistent": Alphabet kept a larger share of each advertising dollar because the Network line, where most revenue is paid away to partners, shrank, not because it negotiated better terms.
Growth in FY2025 was organic. Acquisitions net of cash were US$1.6 billion and the document presents no business combination for the year; Wiz and Intersect close in 2026. Foreign currency is not quantified as a contribution to growth, and hedging swung by 0.1% of revenue.
Gross margin widened from 56.6% in FY2023 to 59.7% in FY2025, because cost of revenues grew more slowly than revenue in both years. Sales and marketing fell to 7.1% of revenue. The lines that did not scale were R&D (+23.8%), driven by compensation including the Waymo charge and by depreciation, and G&A (+51.4%), driven by "expenses related to legal and other matters".
Depreciation is the line to watch. It rose 38.0% to US$21.1 billion, and its share of revenue went from 3.9% in FY2023 to 5.2%. With US$78.6 billion of assets not yet depreciating, the operating leverage visible in gross margin has a cost already on the balance sheet that has not yet reached the income statement.
The question for the income statement is how much of the 32.0% rise in net income came from operations.
| US$ mn | FY2023 | FY2024 | FY2025 | YoY FY2024 | YoY FY2025 |
|---|---|---|---|---|---|
| Revenues | 307,394 | 350,018 | 402,836 | +13.9% | +15.1% |
| Cost of revenues | 133,332 | 146,306 | 162,535 | +9.7% | +11.1% |
| Gross profit | 174,062 | 203,712 | 240,301 | +17.0% | +18.0% |
| Research and development | 45,427 | 49,326 | 61,087 | +8.6% | +23.8% |
| Sales and marketing | 27,917 | 27,808 | 28,693 | −0.4% | +3.2% |
| General and administrative | 16,425 | 14,188 | 21,482 | −13.6% | +51.4% |
| Income from operations | 84,293 | 112,390 | 129,039 | +33.3% | +14.8% |
| Other income (expense), net | 1,424 | 7,425 | 29,787 | n.m. | n.m. |
| Income before taxes | 85,717 | 119,815 | 158,826 | +39.8% | +32.6% |
| Provision for income taxes | 11,922 | 19,697 | 26,656 | +65.2% | +35.3% |
| Net income | 73,795 | 100,118 | 132,170 | +35.7% | +32.0% |
| Diluted EPS (US$) | 5.80 | 8.04 | 10.81 | +38.6% | +34.5% |
| Operating margin | 27.4% | 32.1% | 32.0% | ||
| Net margin | 24.0% | 28.6% | 32.8% |
Operating income grew less than half as fast as net income. The difference sits in other income, which quadrupled to US$29.8 billion, and is dealt with in 3-bis and 6. At operating level, FY2025 carried two legal charges named by the document: the US$3.5 billion EC fine and a US$1.4 billion legal accrual. Adding them back gives operating income of US$133.9 billion and a margin of 33.2%, but the comparison is uneven, because FY2024 legal expense is not quantified and the company expects that it "may continue to incur fines".
| US$ mn | Dec 31, 2024 | Dec 31, 2025 |
|---|---|---|
| Cash, cash equivalents and marketable securities | 95,657 | 126,843 |
| Senior notes (carrying, incl. current portion) | 11,882 | 48,543 |
| Commercial paper | 2,300 | 0 |
| Finance lease liabilities | 1,677 | 2,500 |
| Operating lease liabilities | 14,578 | 15,954 |
| Gross debt (including operating leases) | 30,437 | 66,997 |
| Net cash (cash and securities − gross debt) | 65,220 | 59,846 |
| Total stockholders' equity | 325,084 | 415,265 |
Gross debt includes operating leases throughout; financial debt without them was US$51.0 billion. Alphabet more than doubled its gross debt during FY2025 and still ended the year with more cash than debt. Net cash fell by only US$5.4 billion because most of the new debt was kept as liquidity rather than spent. The reasons are read in 5-bis.
| US$ mn | FY2023 | FY2024 | FY2025 |
|---|---|---|---|
| Cash from operations | 101,746 | 125,299 | 164,713 |
| Capex | (32,251) | (52,535) | (91,447) |
| Free cash flow | 69,495 | 72,764 | 73,266 |
| Stock-based compensation (CF line) | 22,460 | 22,785 | 24,953 |
| Free cash flow after SBC | 47,035 | 49,979 | 48,313 |
| Repurchases of stock (cash) | (61,504) | (62,222) | (45,709) |
| Dividends paid | 0 | (7,363) | (10,049) |
| Net payments related to stock-based award activities | (9,837) | (12,190) | (14,167) |
| Cash taxes paid | 19,164 | 27,353 | 21,526 |
Free cash flow grew 0.7% in FY2025 while net income grew 32.0%. Cash from operations rose US$39.4 billion, and capex absorbed US$38.9 billion of it. Free cash flow converted 55.4% of net income, against 94.2% in FY2023.
The SBC base used throughout is the cash flow figure, US$24.953 billion. Note 13 reports total SBC expense of US$27.1 billion, which includes awards settled in cash or in the stock of Other Bet companies (the Waymo charge among them); the equity statement shows US$25.130 billion. The document does not reconcile the three figures line by line.
Owner earnings are defined here as net income plus non-cash charges, minus non-cash gains, capex and working capital absorbed. SBC is not added back, because net income already carries it as a cost. On that basis owner earnings were US$48.3 billion including working capital, the figure used in the judgment, and US$47.7 billion excluding it. They are below free cash flow by exactly the SBC the cash flow statement adds back. In three years owner earnings moved from US$47.0 billion to US$48.3 billion; net income moved from US$73.8 billion to US$132.2 billion.
Share-based pay has a second, less visible cash cost. Alphabet settles vested awards net of tax and pays the withheld tax in cash. The cash flow statement records US$14.2 billion of "net payments related to stock-based award activities" in financing, outside free cash flow; the equity statement shows US$14.8 billion of tax withholding "and other", a gap the document does not itemize. Note 13 shows 181 million RSUs vested, while the equity statement records 117 million shares issued; the gap is mostly the shares withheld to cover that tax.
| Metric | FY2023 | FY2024 | FY2025 |
|---|---|---|---|
| Gross margin | 56.6% | 58.2% | 59.7% |
| Operating margin | 27.4% | 32.1% | 32.0% |
| Effective tax rate | 13.9% | 16.4% | 16.8% |
| ROIC (year-end invested capital) | not calculable | 42.3% | 37.5% |
| Gross debt / EBITDA | not calculable | 0.24x | 0.45x |
| Net debt / EBITDA | not calculable | −0.51x | −0.40x |
| Interest coverage (EBIT / interest expense) | 273.7x | 419.4x | 175.3x |
| FCF / net income | 94.2% | 72.7% | 55.4% |
| SBC / revenue | 7.3% | 6.5% | 6.2% |
ROIC uses year-end invested capital: equity plus gross debt minus cash and all securities, since the securities earn through other income. NOPAT is taxed at the year's effective rate. On average capital FY2025 ROIC is 42.2%; the 4.8-point gap comes from the US$64.9 billion added in the year, counted in full at year-end and by half on average. FY2023 is not calculable, having no opening balance sheet.
In Alphabet's own terms, interest expense was 0.57% of operating income, and the long-term notes equal 0.36 years of operating income. On gross interest, including US$447 million capitalized into construction, coverage is 109.1x.
The price-based figures below use the US$313 close at December 31, 2025, the last trading day of the fiscal year analysed. That date is chosen so that the price and the earnings share the same reference point. They are not updated to any later price.
| Market measure (price US$313, Dec 31, 2025) | Value |
|---|---|
| Market capitalization (12,097 mn shares) | US$3,786.4 bn |
| Enterprise value (with gross debt incl. leases, minus cash and marketable securities) | US$3,726.5 bn |
| Price / diluted EPS | 28.95x |
| Price / EPS excluding equity-security gains (our construction, see 3-bis) | 33.8x |
| Market capitalization / free cash flow | 51.7x (FCF yield 1.93%) |
| Market capitalization / revenue | 9.40x |
| EV / EBITDA | 24.8x |
| Market capitalization / book equity | 9.12x |
| Owner earnings yield | 1.28% |
| Dividend declared per share / price | 0.27% |
The market capitalization multiplies one price by all three share classes; the document gives no separate price for each class. The distance between the two earnings multiples is the revaluation gains, and the distance between the free cash flow yield and the owner earnings yield is share-based pay.
The notes say how much of each number is management's opinion. The selection follows the auditor first, then size and judgment.
Loss contingencies (the only Critical Audit Matter). The auditor names a single CAM, "Loss Contingencies", because assessing "the probability of a loss, and the estimated amount or range of loss" involved "challenging and subjective auditor judgment". The quantified part is accrued fines and settlements of US$15.6 billion, up US$5.8 billion in the year. That equals 12.1% of FY2025 operating income. For the US ad tech cases, the Digital Markets Act investigations and the private follow-on actions, the document states "we cannot estimate a possible loss"; for the Search judgment and the Android appeal it gives no range. The reasonably possible loss beyond the accrual is not disclosed for any open matter. It would move with the ad tech remedies judgment, the Search appeals and the European Court of Justice ruling on the €4.1 billion Android fine.
Non-marketable equity securities. Stakes in private companies are carried at US$64.1 billion under the "measurement alternative": cost, adjusted only when an observable transaction in the same or a similar security occurs, and written down on impairment. Their FY2025 unrealized net gain of US$21.4 billion equals 13.5% of pre-tax income. The discretion lies in the step from an observed transaction to a value for the security Alphabet actually holds, using unobservable inputs such as "volatility, expected time to exit, risk free rate, and the rights and obligations of the securities we hold". US$45.6 billion of the carrying value was remeasured during the year. Cumulative upward adjustments reached US$44.5 billion against an initial cost of US$28.4 billion; net of US$8.8 billion of downward adjustments, 55.6% of the carrying value is accumulated revaluation rather than money paid. Item 7A warns that the transaction-date value "is not necessarily an indication of the current fair value". A round at a lower price would move it; no round at all leaves it unchanged until an impairment review. No sensitivity is given. The January 2026 gains of US$32.0 billion are "subject to change as we finalize related valuations".
Removing the gains shows what operations earned. Pre-tax income without the equity-security gains was US$134.7 billion, up 16.1% against reported growth of 32.6%. Taxing the gains at the 21% federal statutory rate gives net income of US$113.1 billion and diluted EPS of US$9.25, up 16.4%. This is our construction: the document does not present earnings without the gains, and the statutory rate only approximates the tax attached to them.
Income taxes. The effective rate was 16.8% against the 21.0% federal statutory rate. The main reconciling items are the foreign-derived intangible income deduction (−2.5 points), share-based pay (−1.6 points) and the research credit (−1.3 points). The FDII benefit shrank from −3.8 points, the company's main reason for the higher rate. The valuation allowance rose by US$2.4 billion to US$13.9 billion; it covers California assets, certain loss and credit carryforwards and "net deferred tax assets relating to certain Other Bet companies". Its FY2025 increase cost about US$1.2 billion of tax. A release would be a tax benefit, recorded "if future evidence allows". Unrecognized tax benefits fell to US$11.5 billion after US$2.2 billion of settlements, and the IRS "is currently examining our 2019 through 2021 tax returns". The gap between cash taxes and the provision is read in 3-quater.
Property and equipment useful lives (critical estimate). Plant is 41.4% of total assets, and its cost is governed by the useful life: "generally over a period of six years" for servers and network equipment, about 60% of technical infrastructure. No sensitivity is given, and the document does not say whether lives changed in FY2025. Accumulated depreciation fell from 39.7% to 37.0% of in-service cost, so most of the fleet's depreciation is still ahead. The risk factors state that "expected technology advancements" could "change the period over which we expect to benefit from the asset"; a shorter life would raise depreciation on the whole fleet.
Debt. Face value rose from US$12.0 billion to US$49.1 billion through new dollar and euro notes, all senior unsecured and all fixed-rate except US$500 million. US$36.1 billion matures after 2030, and the document mentions no financial covenants or convertible instruments. Discretion is small; the figure would move with new issuance, which MD&A says the company will "continue to assess".
Share-based compensation. Unrecognized cost on unvested RSUs is US$42.9 billion over a weighted 2.6 years: dilution already contracted, about US$16.5 billion a year before new grants. The Proxy Statement asks to add 200 million Class C shares to a reserve of 534 million that the board calls "insufficient to meet our anticipated needs".
Revenue, deferred revenue and backlog. Deferred revenue rose 43.3% to US$8.6 billion. Remaining performance obligations of US$242.8 billion are "primarily related to Google Cloud", with "just over 50%" expected within 24 months. Their timing depends on "our ability to deliver in accordance with relevant contract terms", which ties the backlog to the capex in 3-quinquies. Network ads are mostly recorded gross, with partner payments in cost of revenues, while Google Play is recorded net; a shrinking Network line lowers reported revenue faster than gross profit.
Goodwill. Goodwill of US$33.4 billion is 5.6% of assets, of which US$24.9 billion sits in Services. The note says only that "goodwill impairments were not material". It gives no test date and no headroom between recoverable and carrying value, so a passed test without disclosed headroom is not informative.
Leases, commitments and guarantees. The largest off-balance-sheet figures are purchase commitments of US$149.1 billion, "mostly related to technical infrastructure and inventory orders", and leases signed but not commenced of US$58.5 billion. Alongside them sit credit derivatives with a US$16.9 billion maximum exposure, backing third-party data center leases for up to 15 years, and US$5.7 billion of guarantees for energy equipment. The discretion lies in classification: the guarantees are carried at a fair value the note calls "not material", and the maximum exposure appears only in the notes. What would move them is "nonperformance or an industry downturn", in which the company says it may "have excess capacity that we cannot easily redeploy".
Related parties, from the Proxy Statement, are small: two hangar licences involving entities affiliated with the founders (about US$8.4 million received and US$1.6 million paid) and US$42.4 million of venture co-investments alongside funds chaired by director John Doerr. The document does not mention defined-benefit pension plans. Subsequent events are the January 2026 gains, a US$16.0 billion Waymo round "the significant majority of which was funded by Alphabet", a US$9.9 billion power agreement and the Search appeals.
The notes push the reading of FY2025 one way. The number most dependent on management's estimates lifted net income above operating income, and the exposures that would hurt in a downturn sit outside the balance sheet.
The balance sheet grew by US$145.0 billion in one year, and the question is what that growth is made of. Five assets exceed 5% of the total: property and equipment, cash and securities, non-marketable securities, receivables and goodwill. On the other side, accrued expenses and long-term debt exceed it besides equity; the shares are in the appendix.
Property and equipment rose from US$171.0 billion to US$246.6 billion net. Assets not yet in service rose 55.3% to US$78.6 billion, 31.9% of net plant: data centers and servers that earn nothing and are not depreciated until ready, a wait that "may extend from months to years".
Cash and marketable securities of US$126.8 billion include US$6.3 billion of listed equities, which are not cash-like, and US$5.6 billion of assets of consolidated VIEs (Waymo among them), "primarily" cash, "that can only be used to settle obligations of these VIEs". Available liquidity is therefore taken as US$121.2 billion, excluding the VIE cash but keeping the listed equities, and net cash on that basis is US$54.2 billion rather than US$59.8 billion.
Non-marketable securities and goodwill are treated in 3-bis, receivables in 3-sexies. Intangible assets are not presented separately. Capitalized software is "not material", so no capitalization policy shifts costs to the balance sheet.
Net deferred tax assets halved to US$8.2 billion, because two liabilities grew: on property (+US$3.3 billion, accelerated tax depreciation) and on investment gains (+US$5.3 billion, tax on unrealized revaluations). Of gross deferred tax assets of US$48.0 billion, US$24.8 billion is research capitalized for tax purposes, and US$13.9 billion is reserved against.
Accrued expenses of US$55.6 billion include fines of US$15.6 billion and accrued capex of US$8.9 billion. A residual "other accrued expenses" line of US$24.6 billion is not broken down. Other long-term liabilities rose 80.0% to US$8.4 billion, also without a breakdown beyond US$2.1 billion of finance leases.
Debt, with leases inside, was itemized in section 3. On year-end balances, days sales outstanding rose from 54.6 to 57.0. Days payables rose from 19.9 to 27.4, or from 11.9 to 13.4 on operating payables (our construction, see A-ter). Inventory days cannot be computed.
Equity rose by US$90.2 billion. Alphabet retires repurchased shares, so buybacks reduce retained earnings directly (US$42.9 billion in FY2025). Return on year-end equity was 31.8%. Adding back the three years of repurchases in the document (US$169.6 billion) gives 22.6%; the correction covers only three years and understates earlier buybacks. Excluding the equity-security gains gives 27.2%. Retained earnings must be read with the payout: Alphabet distributed 84.3% of net income in FY2023 and 42.0% in FY2025, at declared cost. Retained earnings grew fastest in the year the company retained most, a decision rather than a change in profitability.
Of total assets, 21.3% is valued at observable prices or market inputs (cash and marketable securities). Another 11.5% rests on transaction-based remeasurements with management's inputs (private holdings), and 7.1% on management's estimates (goodwill and deferred tax assets). The largest block, plant at 41.4%, is at depreciated cost governed by the useful-life estimate.
Cash from operations rose US$39.4 billion, and the question is how much of that rise the business generated. The rise decomposes into seven pieces:
| Change FY2025 vs FY2024 (US$ mn) | |
|---|---|
| Higher net income | +32,052 |
| Larger non-cash gains on securities removed | −21,949 |
| Higher depreciation | +5,825 |
| Higher SBC add-back | +2,168 |
| Deferred taxes (from −5,257 to +8,348) | +13,605 |
| Working capital (from −8,406 to +618) | +9,024 |
| Other non-cash items | −1,311 |
| Total change in cash from operations | +39,414 |
Two pieces did not come from selling more: deferred taxes and working capital, US$22.6 billion together, more than half the increase. The first coincides with the July 2025 US tax law, whose effect the document places in operating cash flow without quantifying it. In the second, accrued liabilities contributed US$12.9 billion, including fines accrued but unpaid; the EC ad tech fine was secured with bank guarantees "in lieu of cash payment".
Money of third parties held temporarily totals US$17.2 billion, 23.4% of FY2025 free cash flow. It is the increase in deferred revenue, revenue share owed to partners, accrued liabilities and operating payables, itemized in the appendix. Part of it will be paid out: the fines when appeals end, revenue share when partners are paid.
Taxes paid in cash were US$21.5 billion, US$5.1 billion below the provision of US$26.7 billion and above the current provision of US$18.4 billion. The deferred provision swung to a charge of US$8.3 billion. The current provision fell in the year the US tax law of July 4, 2025 allowed immediate research expensing and accelerated depreciation; the document places those effects in operating cash flow without quantifying them, and does not explain why cash paid exceeds the current provision. The tax deferral is real cash kept in FY2025, but it is a timing effect that reverses as the accelerated deductions run out, and the deferred tax liability on property recorded it.
Declared capex does not cover all the investment. On an accrual basis, adding the US$4.8 billion rise in unpaid capex, it was about US$96.2 billion, and leases added a further US$5.7 billion of assets. Leases not yet commenced and purchase commitments are capacity already promised that the capex line does not show.
Without the deferred-tax and working-capital swing, free cash flow, which grew by US$0.5 billion, would have fallen; this is carried into section 9 as a weakness.
This section asks whether the capital Alphabet is adding earns what its existing capital earns. Only two balance sheets are available, so every return on capital below is a single point, not a trend.
Per-share values. On diluted average shares, revenue per share rose from US$24.16 in FY2023 to US$32.94 in FY2025. Net income per share rose from US$5.80 to US$10.81, and equity per share from US$22.27 to US$33.95. Diluted shares fell 3.9% in two years, so every per-share figure grew faster than its total and shareholders were not diluted. Free cash flow per share is the one that barely moved: US$5.46, US$5.85 and US$5.99.
Return on incremental capital. NOPAT rose by US$13.5 billion while invested capital rose by US$64.9 billion, a return of 20.8% on the capital added in FY2025. That is about half the 42.3% the business earned on its FY2024 capital. The capital added includes the US$28.0 billion increase in assets not yet in service, which by construction earn nothing yet. Excluding that increase, the incremental return is 36.5%. The difference between 20.8% and 36.5% measures the lag between paying for capacity and using it, not the quality of the investment. Which of the two figures will hold cannot be told from one year.
Reinvestment rate and implied growth. Alphabet reinvested 66.4% of NOPAT in FY2025 (capex net of depreciation, acquisitions and working capital), up from 51.7% in FY2024. Multiplied by the year-end ROIC of 37.5%, that reinvestment implies operating growth of 24.9%. Multiplied by the incremental return of 20.8%, it implies 13.8%. Operating income actually grew 14.8%, which matches the incremental return, not the average one. Realized growth below the average-return implication fits a company whose new capital is not yet producing. It also fits new capital that earns less than the old. The document cannot separate the two, because it does not say when the assets under construction will enter service.
The two returns combine into one reading. The existing business earns about 40% on its capital, and the company now reinvests two-thirds of its operating profit at, on FY2025 evidence, about half that rate. Alphabet is no longer a business that collects a high return and distributes it: the share of profit distributed has halved since FY2023. The return on what is reinvested will decide whether that change adds value.
Self-financing. Over the three years, free cash flow totalled US$215.5 billion. Buybacks, dividends and tax paid on vested awards took US$223.0 billion. Net new debt was US$32.3 billion, almost all in FY2025: FY2023 was a small net repayment and FY2024 a net US$0.9 billion. Capex was paid from operating cash in every year, so the expansion is self-financed; it is those three uses together that exceeded three-year free cash flow, by US$7.5 billion.
Capital intensity. Capex over three years was US$176.2 billion, 57.6% of cumulative net income. Capex was 2.70 times depreciation in FY2023, 3.43 times in FY2024 and 4.33 times in FY2025: capacity is being expanded, at an accelerating pace, and the company expects "to significantly increase, relative to 2025, our investment in our technical infrastructure".
Operating return on enterprise value. EBIT divided by enterprise value is 3.46% at the US$313 price, and 3.53% if non-marketable securities are also deducted from enterprise value. This is the operating yield a buyer of the whole company would receive before financing and taxes.
The document does not separate maintenance from growth capex. This measure is not reported by the company: we construct it as cash from operations minus SBC minus depreciation, using depreciation as a proxy for maintenance, to estimate what would remain if Alphabet stopped expanding. It was US$118.6 billion, or US$9.70 per diluted share. It is only as good as the proxy: depreciation on a young fleet may understate the cost of replacing servers, and the figure includes the FY2025 tax deferral. At US$313 it represents a yield of 3.13%, against 1.93% on reported free cash flow. Much of the growth spending is outside capex anyway. R&D of US$61.1 billion, all expensed, includes the shared AI model work reported as an Alphabet-level loss of US$16.8 billion, and three years of R&D (US$155.8 billion) compare with a revenue increase of US$95.4 billion.
The question is whether any balance-sheet line is moving ahead of revenue in a way the income statement does not yet show. Each comparison uses year-end balances for FY2024 and FY2025, the only two available.
Inventory against revenue. Not computable: inventory is not disclosed.
Receivables against revenue. Net receivables grew 20.1% against revenue growth of 15.1%, a gap of 5.1 points, which is flagged. The document offers a partial explanation: receivables include amounts "for services performed in advance of the right to invoice", and Cloud, billed largely on consumption, grew 35.8%. It does not say which business those amounts belong to or split them from billed receivables, so a loosening of payment terms cannot be excluded.
Allowance against gross receivables. The allowance for credit losses grew 5.1% while gross receivables grew 19.9%, a gap of 14.8 points. Coverage fell from 1.65% to 1.45%. Had coverage stayed at the FY2024 ratio, the allowance would have been US$1,054 million instead of US$924 million, a difference of US$130 million. This is arithmetic on the document's figures, not an estimate of an error, and it equals 0.08% of pre-tax income. The combined allowance in Schedule II, which adds sales credits charged against revenue, moved the other way: it rose from US$1,636 million to US$2,356 million.
Payables against cost of revenues. Reported accounts payable grew 52.7% against cost of revenues at 11.1%, a gap of 41.6 points. Most of it is capex, as unpaid equipment purchases in payables roughly doubled to US$6.2 billion. On operating payables only (our construction) growth is 25.6%, still 14.5 points above cost of revenues, but payables added only US$0.9 billion to operating cash. The signal is on the capex side, where US$15.1 billion of plant had been received but not paid at year-end.
Capitalized costs against the related expense. Software capitalization is "not material". Capitalized interest rose to US$447 million, but its share of gross interest fell from 42.0% to 37.8%, so it did not hide the rise in interest cost.
Revenue against cash from operations. Cash from operations grew 31.5% against revenue at 15.1%, a favourable gap of 16.4 points. More than half of the rise came from deferred taxes and working capital (3-quater), so the divergence is positive in sign and borrowed in substance; it is carried to section 9.
Four gaps exceed five points. Three are adverse in sign, two of those are partly explained by the document, and none is large relative to earnings. The comparison that matters is the last: operating cash outgrew revenue in a year when free cash flow did not grow.
Alphabet sits between two sector groups: technology and software (Cloud, subscriptions) and an advertising platform that has become capital-intensive. The metrics that govern the model are two pairs, read below as one passage each, followed by the sector metrics the document provides.
The Search pair: paid clicks and cost-per-click. A paid click is a user clicking an ad on Google Search or another Google-owned property (Gmail, Maps, Play). Cost-per-click is the average Alphabet charges the advertiser for that click, defined by the company as "click-driven revenues divided by our total number of paid clicks". Revenue on this line is therefore volume times price, and the two halves move for different reasons. Clicks follow advertiser spending, product and policy changes and "fluctuations in search queries resulting from changes in user adoption and usage". Price follows "advertiser competition for keywords", device, geographic and product mix, and exchange rates. Both halves come from MD&A, Monetization Metrics, only as changes. For FY2025 the pair compounds as 1.06 × 1.07 = 1.134, implied growth of 13.4%, against 13.35% reported for Search & other: the line and the metrics agree. Not every dollar of the line is a click, but the agreement shows volume and price rising together in roughly equal parts. The document does not show how AI features change the pair; it warns only that "we may monetize differently than our historical consumer and enterprise offerings". The net version of the pair is the TAC rate at 20.3%, but the rate on Search alone is not published. Also unpublished: the level of either metric, query volumes, the share of queries from paid access points and revenue from AI-assisted results.
The Cloud pair: revenue and operating margin, governed by the backlog. Google Cloud sells computing capacity, storage, AI models and the Workspace applications, billed by consumption or subscription. The document publishes no units (customers, capacity sold, utilization), so the pair here is revenue and the margin on it. Revenue was US$58.7 billion, up 35.8%, and operating income US$13.9 billion, both from Note 15. The margin was 23.7%, against 5.2% in FY2023. Each additional dollar of FY2025 Cloud revenue brought 50.4 cents of operating income. Inside the segment, employee compensation grew 7.6% and "other costs and expenses" grew 36.9%. The second line includes the allocated "usage costs for technical infrastructure", charged to segments "generally based on usage, headcount, or revenue". The margin expanded because payroll grew far more slowly than revenue, while infrastructure charges grew with it. Those charges are allocations of central costs, and the document does not disclose them in amounts. The balance governing the pair is the revenue backlog of US$242.8 billion, "primarily related to Google Cloud", equal to 4.14 times FY2025 Cloud revenue (our ratio; the backlog is not exclusively Cloud), with just over half due within 24 months. Its conversion depends on "our ability to deliver", that is, on capacity. Not published: Cloud capex or assets, prior-year backlog, its split by product, customer counts and utilization.
The other lines. YouTube ads and subscriptions together added US$11.9 billion of revenue, less than half of Search's increase, and the document gives them no unit metrics. Google Network shrank on impressions (−7%). Other Bets is too small in revenue to need a pair; its weight is in its losses.
Sector metrics. For the technology group, the document provides deferred revenue (US$8.6 billion, +43.3%) and remaining performance obligations, but gross margin only at company level. It gives no recurring-revenue share, renewal rate or margin by product line. On the capacity side, purchase commitments equal 0.92 times FY2025 cost of revenues; with the uncommenced leases they are the equivalent of plant already ordered.
Some of the standard metrics do not apply to Alphabet. Inventory days and the cash conversion cycle cannot be computed, because inventory is not disclosed. Net debt ratios describe a net cash position and say nothing about solvency. Book value per share describes little: a price of 9.12 times book values earning power that the balance sheet does not record. An orders-to-sales ratio is replaced by the backlog above.
The risks below come from Item 1A and Note 10, ordered by how much of the business each could reach. Materialized means a loss already recorded or an adverse ruling already issued; potential means a risk the document describes but has not yet costed.
The two most severe risks are already rulings, not hypotheses. What remains open is their price, which the document declines to estimate.
The comparison is not possible. The document does not state which risk factors were added or removed relative to the prior year, and the FY2024 Form 10-K was not provided. The prior-year risk factors are not reconstructed from memory. Section 1 lists the new items that the comparative columns of this document do reveal.
The question is who decides at Alphabet and with what checks. Control rests on three share classes: Class A with one vote, Class B with ten, Class C with none. According to the Proxy Statement, Larry Page holds 27.4% of total voting power and Sergey Brin 25.3%, and directors and executive officers together hold 54.3%.
The board has ten members, three of them employees (Page, Brin and Pichai), and is chaired by John Hennessy; a Risk and Compliance Committee was added in October 2025. The executive officers are Sundar Pichai (CEO), Anat Ashkenazi (CFO), Ruth Porat (President and Chief Investment Officer), Philipp Schindler (Chief Business Officer) and Kent Walker (Chief Legal Officer).
Management's ownership is disclosed only for voting stock: Pichai holds 227,560 Class A shares, and the other named executives hold none directly. Class C stock, in which equity pay is settled, "is not included in the table", so the document does not show how much of the company management owns. The Proxy states that all directors and named executives met minimum ownership requirements and that pledging and hedging are prohibited. The 10-K discloses new Rule 10b5-1 selling plans by the Chair and by Ruth Porat.
Pay, from the Proxy Statement: Pichai's 2025 reported compensation was US$10.9 million, with no equity grant, because CEO awards follow a three-year cycle. His March 2026 award combines time-based units, performance units tied to total shareholder return relative to S&P 100 companies, and units on Waymo and Wing that vest on the "increase in per unit value" of those companies. The other named executives receive time-based units and performance units on the same relative-TSR measure; their cash bonus program was discontinued in 2025.
The auditor is Ernst & Young, in place since 1999. Fees were US$68.7 million in FY2025, of which US$19.4 million were "all other fees". Related-party transactions are listed in 3-bis and are small.
Governance at Alphabet is a combination of founder control and a professional executive team. With a majority of the vote held by directors and officers, outside shareholders cannot change the board by voting.
The question is what management did with the capital it controlled, judged by the decisions and the numbers they produced.
Capital allocation. Over three years Alphabet generated US$391.8 billion of operating cash and borrowed a net US$32.3 billion. Capex (US$176.2 billion) and buybacks (US$169.4 billion) took most of it; the remaining uses are in the appendix. The composition changed. Capex took 31.7% of operating cash in FY2023 and 55.5% in FY2025. Buybacks fell from 60.4% to 27.8%. This shift is the main decision recorded in the three years. Acquisitions in the comparative years were small, and no impairment of them is reported.
Buybacks. In FY2025 Alphabet retired 240 million shares at a cost of US$45.4 billion, an average of US$189.16 per share. That is 39.6% below the US$313 year-end close and 17.5 times the year's diluted EPS. Against free cash flow per share, the company paid 21.6 times in FY2023, 28.0 times in FY2024 and 31.6 times in FY2025 (our construction). Spending held at about US$62 billion while the multiple rose to 28.0, then fell 26.8% when it reached 31.6. In the fourth quarter purchases averaged about US$263 in October and November, and stopped in December. The cost in the equity statement, used here, is US$0.3 billion below the US$45.7 billion of cash paid; Note 11 says the cost includes unsettled repurchases and the document does not itemize the gap. The year's free cash flow covered buybacks, dividends and tax on vested awards with US$3.3 billion to spare, so the new debt was not needed for them.
Net dilution. Shares issued for compensation were 117 million and shares retired 240 million, a net reduction of 123 million in FY2025. Over three years shares outstanding fell 5.9%. The buyback is returning capital, not only offsetting dilution, with the caveat from section 3 that the tax paid in cash on vested awards is a further cost outside free cash flow.
Accounting aggressiveness. The 10-K presents no adjusted or non-GAAP measures, so there are no exclusions to add up. Fines stay inside operating income, which matters because they recur: the document records EC fines decided in 2017, 2018, 2019 and 2025, and settlements in 2023 and 2025. What flatters FY2025 is not an exclusion but an inclusion, the securities gains inside net income.
Governance signals in the document. Internal control was judged effective, there were no restatements, the auditor has not changed since 1999, and related-party dealings are small. Late Section 16 reports concerned minor acquisitions of units by seven directors and one executive.
The decisions of FY2025, read together. In one year Alphabet issued long-dated dollar and euro notes "for general corporate purposes" (Note 6) and cut buybacks (Note 11). It signed an all-cash acquisition of Wiz for US$32.0 billion and agreed to buy Intersect, "which provides data center and energy infrastructure solutions" (Note 8). It began to backstop third-party data center leases and to guarantee energy equipment (Notes 3 and 10), signed US$58.5 billion of data center leases starting from 2026 (Note 4), and in February 2026 funded "the significant majority" of a US$16.0 billion Waymo round (Note 5). Cash and securities rose by US$31.2 billion, about the net borrowing and about the price of Wiz. Together these choices move cash from buybacks to computing capacity and to the power and land it needs, and they carry part of that capacity through leases, backstops and an acquired developer rather than through capex. Next year's balance sheet will show the consequences: Wiz goodwill, Intersect's debt, new lease liabilities, a larger Waymo noncontrolling interest, and any backstop called. The document states no purpose beyond "general corporate purposes".
Remuneration. The performance measures in the Proxy Statement are relative total shareholder return for performance units and the change in per-unit value of Waymo and Wing. None of the measures is built on return on capital, free cash flow or capex, the three quantities that the FY2025 decisions moved most. With no adjusted figures, there is no exclusion for pay measures to mirror.
Does the document admit an error, a closed project, a missed target, a write-down or a forecast cut? It reports adverse outcomes when a court or the numbers impose them: "the US District Court for the District of Columbia ruled against Google"; "Google Network revenues decreased $567 million"; devices "have had, and may continue to have, an adverse effect on our consolidated margins". It contains no admission of a decision of management's own that went wrong: no project closed, no target missed, no investment written down with an explanation. On the antitrust matters it adds "we believe we have strong arguments against these open claims". The absence of admissions is an observation, not an accusation.
The question is which accounting choices move FY2025 profit, and in which direction.
Principles changed little: the new tax disclosure standard was adopted, and standards on expense disaggregation (2027) and internal-use software (2028) are pending, the second able to change how much development is capitalized. "Certain amounts in prior periods have been reclassified", without saying which. The single Critical Audit Matter is treated in 3-bis.
Three choices move profit. The measurement alternative turns other investors' funding rounds into Alphabet's income: US$21.4 billion in FY2025, after US$3.4 billion in FY2024 and a loss in FY2023. The useful life of servers spreads the cost of the build, with no sensitivity given. Awards in Other Bet stock are remeasured through operating expense, so a higher Waymo valuation produced a US$2.1 billion compensation charge. Rising private valuations thus lower operating income and raise net income at the same time; in FY2025 the net effect was positive and large.
Net income exceeded free cash flow by US$58.9 billion. Capex above depreciation (US$70.3 billion) and non-cash gains (US$24.6 billion) explain it, partly offset by the SBC add-back, the tax deferral and working capital. The gap has widened in each year since FY2023, mainly through capex.
The tax rate reconciliation is read in 3-bis. The quality of FY2025 earnings is high at the operating line, which contains every recurring cost including fines, and lower below it, where 15.2% of pre-tax income is gains on equity securities, mostly unrealized.
The question is which developments already in the document could raise reported results. Each item below is anchored to a statement or figure in the 10-K, and none extends beyond what the document says.
Each catalyst has a cost already visible in the document. Backlog needs capacity, capacity brings depreciation, Wiz will be paid in cash, the tax deferral reverses, and revaluations can reverse.
A comparison between what Alphabet declared a year earlier and what happened is not possible. This document contains no prior-year targets in its comparative columns or notes, and the FY2024 Form 10-K was not provided. The 10-K publishes no numerical margin, revenue or capex targets. The only commitments it reports as carried out are the dividend increase of 5% to US$0.21 per quarter in April 2025 and a second US$70.0 billion repurchase authorization; US$69.5 billion remained available at year-end.
Next year's 10-K will be able to check four statements made here: capex will "significantly increase, relative to 2025"; Wiz will close in 2026 and Intersect in the first half; just over half the backlog will become revenue within 24 months; and the signed leases "will commence between 2026 and 2031".
Suppose that three years from now an investment in Alphabet at US$313 has produced a permanent loss of capital. Three mechanisms in the document could have caused it.
The capacity does not earn its cost. Mechanism: the incremental return of 20.8% does not rise toward the existing 42.3% as assets enter service, and part of the commitments, leases and backstops becomes "excess capacity that we cannot easily redeploy". The figure today is capex at 4.33 times depreciation, with 31.9% of net plant not yet in service. The signal: depreciation growing faster than revenue as construction moves into service, with a Cloud margin that stops widening.
The Search pair breaks. Mechanism: court-ordered data sharing and syndication, or a shift of queries to AI interfaces that monetize less, stops clicks and price from rising together on 55.7% of revenue. The figure today is paid clicks +6% and cost-per-click +7%, with a TAC rate of 20.3%. The signal: either metric turning negative, or a rising TAC rate.
The price rested on earnings that included revaluations. Mechanism: the 28.95 times reported EPS paid at US$313 included securities gains, and more followed in January 2026. Private valuations fall, part of the US$44.5 billion of upward adjustments reverses, and reported earnings fall with no change in operations. The figure today is gains equal to 15.2% of pre-tax income. The signal: "cumulative downward adjustments (including impairments)" in Note 3, US$8.8 billion at year-end.
No probabilities are assigned. The three mechanisms are not independent: the first and the third both depend on the value of AI-related assets, inside and outside the balance sheet.
What would remain for the shareholder if growth stopped? Tangible equity, after removing US$33.4 billion of goodwill, is US$381.9 billion, or US$31.59 per share on the 12,088 million shares at year-end. Intangible assets other than goodwill are not separately disclosed, so this figure removes goodwill only. Net working capital is US$103.3 billion, with a current ratio of 2.01. Cash and securities are US$10.49 per share, and US$4.95 per share after deducting all gross debt including leases. The split of total assets by valuation basis is the one that closes 3-ter.
If revenue stopped growing and margins returned to the average of the three years, the cash and the working capital would protect the shareholder; the private holdings and the plant built for growth that did not come would not, and at US$313 the protected part, US$4.95 of net cash per share, is 1.6% of the price.
The three questions set in 1-bis can now be answered. What does the capital being committed earn? On FY2025 evidence, about half of what the existing capital earns, or about 86% of it if the plant not yet in service is set aside (3-quinquies). Realized operating growth matches the lower figure, and the company has committed more capital outside the balance sheet than it spent in the year. The answer rests on two balance sheets, and the document lacks the segment assets, capacity units and in-service dates that would make it a trend.
How much of FY2025 profit is operating profit? About 85% of pre-tax income. Without the securities gains, earnings grew about 16% rather than 32%, and owner earnings, which treat share-based pay as a cost, were flat for three years.
Is the Search engine still growing by volume and price together? Yes in FY2025, in roughly equal parts. The document says nothing about the effect of AI interfaces on either half, and the court remedies are not costed.
What the price implies, in arithmetic only. At US$313 the buyer pays 33.8 times earnings without the gains, and receives a 1.93% free cash flow yield. The cash yield is low because free cash flow carries US$70.3 billion of capex above depreciation; on our maintenance proxy it is 3.13%. The gap between those two yields is the expansion, and the price rests on that expansion earning returns closer to the existing capital's than to the FY2025 incremental return.
The reading holds if assets placed in service lift NOPAT faster than invested capital, which would appear as a rising incremental return and a Cloud margin that keeps widening. It falls if depreciation grows faster than revenue once the US$78.6 billion of assets under construction is in service, if the Search pair weakens, or if the January 2026 gains reverse in later notes.
The document does not split capex into maintenance and expansion, or by segment. It does not say when the assets under construction will enter service or what the backlog was a year earlier. It gives no levels for clicks or price and no revenue from AI features, no cost for the Search remedies and no loss range for ad tech, and no Class C holdings of management. Each answer would change the reading of at least one decisive question.
Their movement over the next twelve months will show whether the capital committed in FY2025 is earning what the price assumes.
Report based on Alphabet Inc.'s Form 10-K for the fiscal year ended December 31, 2025 (filed with the SEC on February 5, 2026) and, for ownership, compensation and related parties, on the 2026 Proxy Statement. Multiples use the closing price of US$313 at December 31, 2025, and 12,097 million shares outstanding at January 28, 2026, both supplied with the request. This report is for analysis purposes only and does not constitute investment advice.
All amounts in US$ millions unless stated. Inputs are in the data tables (A, A-bis, A-ter).
Section 1
Section 1-bis
Section 2
Section 3
Section 3-bis
Section 3-ter
Section 3-quater
Section 3-quinquies
Section 3-sexies
Section 3-septies
Section 5-bis
Section 6
Section 8-ter
Section 9
The full data table behind this report — every line item with its source in the filing and the arithmetic checks — is published separately: Alphabet Inc. FY2025 data table.