Google Cloud grew 103% in the third quarter and its backlog reached $561bn. Search grew 12%, about three points short of expectations, and on 29 October the shares had their worst day since listing in 2004. On consensus revenue and a capex programme that absorbs almost all of the cash until 2029, the shares are worth $245.08 to $320.06, most likely $277.04. At $279.84 the price pays for the most likely path and for nothing beyond it. We rate the shares fairly valued.
Cloud ran ahead of expectations; Search fell behind them. Google Cloud revenue rose 103% to $30.7bn at a 36.2% operating margin, with TPU system sales a meaningful contributor, and the Cloud backlog reached $560.8bn, just over 55% of it due within 24 months. Search & other grew 12%, after 19% in the first quarter and 17% in the second, against about 15% expected; management named currency first, then retail and media advertisers, then the lower monetisation of fast-growing AI Mode queries. Our base case follows consensus revenue to 2028, about $631bn in 2027 and $770bn in 2028, with less from Services and much more from Cloud.
The cash gap is wider and lasts longer. Free cash flow was −$9.1bn in the quarter after $51.4bn of capital spending and a $5.2bn payment for an EU fine, and $19.7bn over twelve months. Debt reached $126.1bn and $12.4bn of stock was sold through the at-the-market programme. Capital spending will “increase significantly” in 2027; we use about $300bn, in line with reported analyst expectations. On that path, 2026–2030 capital spending of $1.51tn absorbs 88% of core EBITDA and leaves $46bn of free cash flow in five years.
The price now pays for our base case and little more. After the 12.6% fall on 29 October, $279.84 sits 1.0% above our $277.04 base case, inside the $245.08–$320.06 band and 1.3% below the $283.39 probability-weighted value. A 9.44% discount rate, or steady-state capital spending of 19.7% of revenue, reproduces it. Cash flows to 2030 contribute $1.01 a share; the rest is the 2030s, the terminal value and the stakes.
At $279.84 Alphabet’s market value is about $3.46tn on 12,348m shares, counting the mandatory convertible preferred as converted, and its enterprise value about $3.41tn after $174.7bn of cash and debt securities and $126.1bn of debt. That is 18.6 times consensus earnings for 2027, below Microsoft and Amazon, and about 175 times a twelve-month free cash flow of $19.7bn. Our base case takes consensus revenue to 2028, gives Google Services 9% growth in 2027 fading to 3%, lets Google Cloud grow to $465bn by 2030 at a 57% EBITDA margin in the long run, and charges capital spending of $300bn in 2027 and $330–340bn a year to 2030 before it eases to 20% of revenue. That path is worth $277.04 a share, 1.0% below the price.
Run the model in Section 02 backwards and the price needs, one at a time with everything else at the base case, a 9.44% discount rate instead of 9.5%; or steady-state capital spending of 19.7% of revenue instead of 20%; or 2027–2030 capital spending 3% lower; or Cloud growth 1% faster. Put plainly, the market is pricing our base case almost exactly: it is not paying for faster Search monetisation, lower capital spending or the listing value of the Anthropic stake, and it is not discounting the bear case’s disruption either. The price sits inside our $245.08–$320.06 band, 13% below its top.
The signature measure for Alphabet is when the cash arrives. The first bar takes the share price apart; the second is the Cloud backlog by when it becomes revenue; the third sets the backlog due within two years against our base-case Cloud revenue over the same period.
Each segment of the first bar is a present value at 9.5%. Free cash flow is negative in 2026–2028 and positive from 2029, so the five years to 2030 net to almost nothing: capital spending of $1.51tn against $1.71tn of core EBITDA. The stakes are after tax on the embedded gain; net cash is after debt and $13.8bn of accrued fines and settlements. Backlog and its timing are from the Form 10-Q, note 2 (“just over 55%” within 24 months; drawn at 55%). The backlog includes $16.8bn under a multi-year agreement with Crux AI, a joint venture with Blackstone in which Alphabet holds a minority stake. Anthropic’s registration statement (26 October 2026) discloses a commitment to spend at least about $114.5bn with Google between April 2026 and July 2033; Alphabet does not say how much of it is in the backlog. About two-thirds of our base-case Cloud revenue over the next two years is already contracted; almost none of the value is earned before 2031.
The 2027 capital-spending guide, due with the fourth-quarter results in early February. We use about $300bn; every 10% on 2027–2030 spending moves the value by about $9.27 a share. The second test is Search: whether growth re-accelerates once the late-2025 surge is lapped in the first half of 2027, or keeps slowing while usage grows.
The 10-year Treasury yield was 5.05% on 28 October. We use a 9.5% discount rate in every case: a beta of about 1 and an equity risk premium of about 4.5%, with debt too small to move the blend. Search disruption, capex returns and customer concentration are carried in the scenarios and their weights rather than added to the rate. The band runs the rate from 9% to 10%; at 10.0% the base case is worth $253.96.
2026–30 capital spending against of core EBITDA · free cash flow in 2027 and in 2030 · terminal value is of core enterprise value · shares, with the preferred as converted.
This is the model that produced every figure in this report, reduced to its ten moving parts. It is an unlevered discounted cash flow in sum-of-the-parts form. Google Services and Google Cloud are grown on separate paths to 2040; each earns an EBITDA margin after stock pay; Alphabet-level AI research and a small Other Bets cost are deducted as a share of revenue. Capital spending is a dollar plan to 2030 that glides to a steady share of revenue by 2033, and depreciation follows it by vintage: 65% of each year’s spending is servers and network equipment over six years, 35% buildings over 25 years. Cash tax is 17% of operating profit less the deferral from full first-year expensing of US equipment. Waymo is excluded from the cash flows and valued at its last funding round. The valuation date is 31 October 2026; 2026 is part-reported, so only the fourth-quarter flow is discounted, each later year at its mid-point, and the Gordon terminal value at 3% on the 2040 cash flow from the end of 2040. On base settings the terminal value is 56% of core enterprise value. The scenario cards load whole published paths, and the sliders bend the loaded path rather than replace it, so each card reproduces its published value exactly.
Why these inputs. Services growth of 9% in 2027 (assumed), below the 13.6% of the first nine months because comparisons stay hard to mid-2027 and AI Mode monetisation is still catching up; it fades to 3% by 2039. Cloud growth of 86% in 2027 and 50% in 2028 (consensus-shaped). Consensus revenue of about $631bn and $770bn does not split the segments, so we give the slowdown to Services and the rest to Cloud; TPU system sales “ramp meaningfully” in 2027, and the backlog covers about two-thirds of our Cloud revenue over the next 24 months. Growth slows to 19% by 2030 and 3% by 2040. Long-run EBITDA margins of 48.25% for Services and 57% for Cloud (assumed), against about 46.1% and 46.8% in the first nine months on our depreciation allocation; Cloud’s path treats TPU hardware as no less profitable than services over time, and a hardware-diluted path (54%) would give $259.31. Alphabet-level AI spending of 4.6% of revenue in 2026 easing to 3.75% (assumed), against about 4.4% in the first nine months and 5.0% in the third quarter. Capital spending of $200bn in 2026, $300bn in 2027 and $330–340bn a year to 2030 (assumed): the 2026 guide midpoint, then reported analyst expectations for 2027 after management said spending would “increase significantly”. Steady-state capital spending of 20% of revenue from 2033 (assumed), about double the pre-AI 10% of 2023, for a business in which Cloud becomes more than half of revenue. Discount rate 9.5% and terminal growth 3.0% (assumed), the same in every case.
What the controls hold fixed. The 2026 inputs: nine months of reported segment revenue, a fourth quarter with Services up 9% and Cloud revenue of $37.0bn, and the nine-month reported free cash flow. Working capital at 2% of the change in revenue plus a TPU inventory build of $10bn in 2026, $8bn in 2027 and $4bn in 2028. The SpaceX shares at the 30 October price, other listed holdings and Waymo at their latest marks, each after 22% tax on the gain. The $13.8bn of accrued fines and settlements is deducted as debt. The fair-value band in Section 03 is this model run across the discount rate and terminal growth, and every cell of that grid is reproducible from the two controls.
Derivations, base case. 2026 revenue: Services $280.5bn for nine months plus a fourth quarter of $104.5bn ($95.9bn a year earlier, up 9%); Cloud $75.5bn plus $37.0bn. Fourth-quarter free cash flow −$24.0bn: the modelled year less the nine-month reported $−4.8bn, adjusted for an estimated $4.5bn of Waymo spending and the $5.2bn EU fine paid in July, discounted half a month. Depreciation allocation for margins: about 45% Services, 40% Cloud, 15% Alphabet-level (assumed; segment depreciation is not disclosed). Shares 12,348m: 12,294m on the cover of the Form 10-Q (21 October 2026) plus 53.8m for the 19m mandatory convertible preferred shares at their maximum conversion rates, which apply below about $352. Net cash $29.5bn: $174.7bn of cash and debt securities (the $265.1bn line less $90.4bn of marketable equity) less $126.1bn of debt, $13.8bn of accrued fines and settlements, an estimated $2.5bn of Waymo cash owned by its outside investors and $2.8bn for preferred dividends above the as-converted common dividend to 2029. Stakes: SpaceX 551.2m shares × $134.72 = $74.3bn, $58.1bn after tax; other listed equity $6.4bn; non-marketable securities $117.7bn ($128.2bn carrying, $48.6bn cost, plus $7.1bn of equity-method holdings); Waymo $77.4bn (≈80% of the $110bn pre-money value of February, an estimated $40bn invested).
The number above is not a fact about Alphabet. It is what these assumptions produce, and the one that decides it is not in any filing: how long capital spending absorbs the cash and what it earns. The price sits inside every cell of the band; the rating changes to overvalued above $320.06 and to undervalued below $245.08. Move the capital-spending and Cloud settings until they are yours, and use the number that falls out.
Each card loads its published case into the model above and prints its value: $105.83, $277.04, $426.00 and $515.32. All four use the same 9.5% discount rate and 3.0% terminal growth, so they differ only in what Alphabet sells, at what margin and with how much capital spending; disruption, concentration and capex risk sit in the cases and the weights, not in the rate. Weighting the four at 25 / 45 / 25 / 5 gives $283.39.
| Input | Bear (25%) | Base (45%) | Bull (25%) | Super-bull (5%) |
|---|---|---|---|---|
| Services growth, 2027 / 2030 / 2035 | 5% / 3% / 3% | 9% / 6.5% / 4% | 10.5% / 8.5% / 6% | as bull |
| Cloud growth, 2027 / 2028 / 2030 | 55% / 25% / 11% | 86% / 50% / 19% | 95% / 58% / 24% | 100% / 65% / 30% |
| Cloud revenue, 2030 | $276bn | $465bn | $559bn | $666bn |
| Services EBITDA margin, long run | 43% | 48.25% | 50% | 50% |
| Cloud EBITDA margin, long run | 48% | 57% | 60% | 61% |
| Alphabet-level AI spending, % of revenue | 4.6% → 4.4% | 4.6% → 3.75% | 4.6% → 3.75% | 4.6% → 3.75% |
| Consolidated EBITDA margin, 2035 | 40.5% | 49.0% | 51.9% | 53.3% |
| Capital spending, 2027 / 2030 | $285bn / $220bn | $300bn / $340bn | $310bn / $380bn | $320bn / $440bn |
| Steady-state capital spending, from 2033 | 24% | 20% | 18% | 19% |
| Free cash flow, 2027 / 2030 | −$71bn / $47bn | −$52bn / $88bn | −$52bn / $127bn | −$59bn / $127bn |
| Common to all | 2026 from nine months reported; discount rate 9.5% and terminal growth 3.0%; book tax 17% less the first-year expensing deferral; stock pay inside the margins as a cost; depreciation by vintage; valued at 31 October 2026 from the 30 September 2026 balance sheet; stakes at latest marks after tax; accrued fines deducted as debt | |||
| Value per share | $105.83 | $277.04 | $426.00 | $515.32 |
| $ billions, base case | 2026 | 2027 | 2028 | 2029 | 2030 | 2033 | 2040 |
|---|---|---|---|---|---|---|---|
| Google Services revenue | 385 | 420 | 453 | 485 | 516 | 606 | 775 |
| Google Cloud revenue | 113 | 210 | 315 | 391 | 465 | 659 | 943 |
| Revenue, excluding Other Bets | 497 | 629 | 768 | 875 | 981 | 1,265 | 1,719 |
| EBITDA after stock pay | 204 | 272 | 350 | 412 | 470 | 618 | 844 |
| EBITDA margin | 41.0% | 43.3% | 45.5% | 47.1% | 47.9% | 48.9% | 49.1% |
| Depreciation | 30 | 52 | 85 | 121 | 159 | 239 | 258 |
| Capital spending | (200) | (300) | (330) | (340) | (340) | (253) | (344) |
| Capital spending, % of revenue | 40% | 48% | 43% | 39% | 35% | 20% | 20% |
| Cash-tax deferral | 18.3 | 24.3 | 22.3 | 17.9 | 12.5 | (6.3) | 3.5 |
| Free cash flow to the firm | (19) | (52) | (10) | 39 | 88 | 293 | 404 |
Sonde estimates. 2026 is the full year; only the fourth-quarter flow of −$24.0bn is discounted. Street-basis EPS implied by the base case, after Waymo losses: 2027 $14.36 and 2028 $17.32, against consensus of about $15.08 and $18.15; the gap is our faster depreciation, which does not affect cash flow. Capital spending exceeds depreciation by about $1.06tn over 2026–2030; the cash-tax deferral is worth about $95bn over the same years.
| Base case | $bn | Note |
|---|---|---|
| Present value of cash flows, Q4 2026–2040 | 1,364 | Mid-year convention, the fourth quarter as a stub |
| Present value of terminal value | 1,768 | 56% of core enterprise value; 2041 free cash flow $416bn at 9.5% less 3.0% |
| Core enterprise value | 3,132 | |
| + Cash and debt securities | 174.7 | $265.1bn less $90.4bn of marketable equity, 30 September 2026 |
| − Debt | (126.1) | Long-term debt $125.1bn plus the $1.0bn current portion |
| − Accrued fines and settlements | (13.8) | Form 10-Q, note 7 |
| − Other debt-like items | (5.3) | Outside investors’ share of Waymo cash ($2.5bn, estimated); preferred dividends above the as-converted common dividend ($2.8bn) |
| + SpaceX, after tax | 58.1 | 551.2m shares at $134.72, 22% tax on the gain; $75.2bn of the shares under short-term sale restrictions at 30 September |
| + Other listed equity, after tax | 6.4 | $7.2bn at 30 September, taxed on half |
| + Non-marketable securities incl. Anthropic, after tax | 117.7 | $128.2bn carrying (mostly one private company) over a $48.6bn cost, plus $7.1bn of equity-method holdings |
| + Waymo, after tax | 77.4 | ≈80% of the $110bn pre-money value of February 2026; an estimated $40bn invested |
| Equity value | 3,421 | Components rounded |
| Shares, millions | 12,348 | 12,294m common plus the mandatory preferred as converted |
| Value per share | $277.04 |
The stakes and net cash are worth about $23.42 a share, 8% of the price. At $279.84 the market is paying about $3.17tn for the operating business against our $3.13tn.
The grid holds the base case fixed and varies the discount rate against terminal growth. Its corners are the published fair-value band, $245.08–$320.06: a WACC of 9–10% by terminal growth of 2.5–3.5%.
| WACC ↓ / terminal growth → | 2.5% | 3.0% | 3.5% |
|---|---|---|---|
| 9.0% | $290.67 | $304.14 | $320.06 |
| 9.5% base | $266.17 | $277.04 | $289.72 |
| 10.0% | $245.08 | $253.96 | $264.19 |
The $279.84 price lies inside the band, 14% above its bottom and 13% below its top: fairly valued. The rating would change to overvalued above $320.06 and to undervalued below $245.08. Run across the same grid, the probability-weighted value spans $250.27–$327.96. The second table holds the rates at the base case and moves one input at a time:
| Change to the base case, all else held | Value | Change |
|---|---|---|
| Services growth 1 point higher every year from 2027 | $289.61 | +5% |
| Services growth 1 point lower every year from 2027 | $265.66 | −4% |
| Cloud growth 10% faster, 2027–2035 | $307.88 | +11% |
| Cloud growth 10% slower, 2027–2035 | $249.76 | −10% |
| Both segment margins 1 point higher from 2027 | $288.06 | +4% |
| Both segment margins 1 point lower from 2027 | $266.02 | −4% |
| Cloud margin on a hardware-diluted path (54% long run) | $259.31 | −6% |
| Capex 2027–2030 10% lower | $286.31 | +3% |
| Capex 2027–2030 10% higher | $267.77 | −3% |
| Steady-state capex 18% of revenue | $295.70 | +7% |
| Steady-state capex 22% of revenue | $258.38 | −7% |
| Anthropic stake at $225bn before tax (a $1.5tn listing, ~15%) | $283.67 | +2% |
| Anthropic stake at $300bn before tax (a $2tn listing, ~15%) | $288.41 | +4% |
| Accrued fines and settlements not deducted | $278.16 | +0% |
Modal base case $277.04, 1.0% below the price. Fair-value band $245.08–$320.06, the WACC 9–10% by terminal growth 2.5–3.5% grid, which anchors the rating; the price is inside it: fairly valued. Probability-weighted value $283.39 across the four cases at our own weights, 1.3% above the price and a little above the base case: the bull cases sit further from the base ($148.96 for the bull) than the bear does ($171.21 below), but the bear’s weight is the same, so the skew is mild.
| Single input changed, all else at base | Price implies | Our base |
|---|---|---|
| Discount rate | 9.44% | 9.5% |
| Steady-state capital spending, % of revenue | 19.7% | 20% |
| Capital spending, 2027–2030 | 3% lower | $1.31tn |
| Cloud growth, 2027–2035 | 1% faster; 2035 Cloud ≈$776bn | $765bn |
| Both segment margins, every year from 2027 | +0.3 points | 49.0% group, 2035 |
| Services growth, every year from 2027 | +0.2 points | 9% → 3% |
Every lever lands next to our own setting: the market is pricing the base case. What moves the rating is a change in the inputs, not a re-reading of them. The band top of $320.06 needs a 9.0% rate with 3.5% terminal growth, steady-state capital spending of about 15% of revenue, or Cloud growth about 14% faster than our consensus path every year to 2035. The band bottom of $245.08 needs a 10% rate with 2.5% growth, steady-state capital spending of about 23% of revenue, Cloud growth about 12% slower, Services growth three points a year lower, or a third more capital spending in 2027–2030.
| Scenario | Value | Against price | Weight | Contribution |
|---|---|---|---|---|
| Bear | $105.83 | −62% | 25% | $26.46 |
| Base | $277.04 | −1% | 45% | $124.67 |
| Bull | $426.00 | +52% | 25% | $106.50 |
| Super-bull | $515.32 | +84% | 5% | $25.77 |
| Probability-weighted | $283.39 | +1% | 100% | $283.39 |
We weight the bear at 25% because this is the first quarter in which Search usage and Search revenue clearly diverged, and because capital spending of this size has never been tested through a downturn in demand. We weight it no higher because the Cloud backlog is contracted, rising and increasingly near-term, and management attributes most of the Search slowdown to currency and comparisons. Weights are our judgement, not a model output.
| Risk | Priced in | Not also in |
|---|---|---|
| AI disruption of Search | The bear case (Services growth 3% from 2029, margin 43%) and its 25% weight; base Services growth already 9% in 2027 | The discount rate |
| Capital spending that earns too little | The capex path and steady-state intensity in each case (24% bear, 20% base, 18% bull) | The discount rate or terminal growth |
| TPU hardware margins | The bear Cloud margin (48%); the hardware-diluted alternative is shown, not charged | Base-case revenue |
| Regulation and fines | Remedies in force are in the reported numbers; $13.8bn of accrued fines deducted as debt; the bear case for escalation | A separate haircut |
| AI customer concentration (Anthropic, Crux AI) | The bear Cloud path | A haircut to the backlog |
| Marks on the stakes | Latest marks after tax; SpaceX at the market price | A discount to the marks |
| Dilution | Stock pay charged as a cost; the preferred as converted | An add-back |
| Equity-market and rate risk | The 9.5% discount rate and the 9–10% grid | The cash flows |
| Company | Mkt value | EV | Growth, latest quarter | Op. margin, latest quarter | Fwd P/E | PEG | EV / revenue | EV / EBITDA | P / FCF |
|---|---|---|---|---|---|---|---|---|---|
| Alphabet $279.84 | $3.46tn | $3.41tn | +25% | 31.8% | 18.6× | 0.91 | 7.2× | 18.4× | 175× |
| Microsoft $521.91 | $3.88tn | $3.85tn | +19% | 49.8% | 26.1× | 1.30 | 11.1× | 19.1× | 65× |
| Amazon $240.35 | $2.60tn | $2.67tn | +12% | 13.9% | 22.5× | 0.69 | 3.3× | 14.5× | n/m |
| Meta $614.99 | $1.57tn | $1.57tn | +26% | 20.1% | 18.0× | 1.04 | 6.5× | 15.1× | 55× |
| Oracle $130.67 | $0.40tn | $0.48tn | +30% | 34.8% | 16.1× | n/m | 6.7× | 14.9× | n/m |
All five at their closing prices on 30 October 2026. Market value is price × the share count on the cover of each company’s latest filing (Alphabet 12,294m at 21 October plus 53.8m for the preferred as converted; Microsoft 7,426m at 22 October; Amazon 10,813m at 21 October; Meta 2,556m Class A and B at 16 October; Oracle 3,024m at 7 September). Enterprise value adds debt and deducts cash and marketable debt securities at each company’s latest quarter-end (Oracle: 31 August); lease liabilities and equity stakes are left out for all five, so Alphabet’s $90.4bn of marketable equity, mostly SpaceX, is not counted as cash. Trailing figures are GAAP and cover the latest twelve months; trailing free cash flow was $19.7bn for Alphabet, $60.0bn for Microsoft, $28.6bn for Meta, −$46.9bn for Amazon and −$28.7bn for Oracle, so P/FCF is not meaningful for the last two. Alphabet’s trailing GAAP P/E of 14.6× includes large gains on its stakes and overstates earnings power, so it is left out. Forward P/E uses consensus EPS for the next fiscal year from a secondary aggregation at 30 October (Alphabet 2027 $15.08; Microsoft fiscal 2027 $19.97; Amazon 2027 $10.67; Meta 2027 $34.14), approximate, and Oracle’s own fiscal 2027 guidance of $8.10, reaffirmed on 28 October; PEG divides it by consensus growth in the following year. EBITDA is operating income plus depreciation and amortisation from the cash flow statement (for Alphabet, depreciation of property and equipment). Meta’s September-quarter margin includes large legal accruals.
Cheap, fair or expensive? At the cheap end on earnings, the most expensive on cash. At 18.6 times next year’s consensus earnings Alphabet trades below Microsoft (26.1×) and Amazon (22.5×), in line with Meta (18.0×) and above Oracle (16.1×). On enterprise value to trailing EBITDA it sits at 18.4×, just below Microsoft. On price to free cash flow it is by far the most expensive of those still generating any, at about 175×. The gap between the two views is the capital spending, which earnings per share do not charge until it depreciates. One sell-side sum-of-the-parts published after the results values core Search at about ten times earnings, below the energy sector, so the earnings discount is already compensating for Search risk; our cash flow model charges that risk only in the bear case, not again in the rate. On the market’s own earnings framing, consensus 2027 EPS of $15.08 at the peer range of about 18–26 times gives roughly $271–394; our value sits near the bottom of that range because it charges capital spending when it is spent.
| What would expand the multiple | What would compress it |
|---|---|
| 2027 capital spending guided at or below about $300bn with Cloud growth intact | 2027 spending well above $300bn without a matching revenue upgrade |
| Search growth re-accelerating as comparisons ease and AI Mode ads scale | Search slipping toward single digits while usage keeps rising |
| Disclosure of TPU system margins at or above Cloud’s | Evidence that TPU systems are low-margin hardware resale |
| An Anthropic listing at or above the valuations discussed | A listing well below them, or a falling SpaceX price |
| Free cash flow turning up before 2029 | Further equity beyond the at-the-market programme |
Steady-state capital spending is the valuation and cannot be observed yet; two points of revenue either way move the base case by about $18.66 a share. 2027 capital spending is not guided; our $300bn follows reported analyst expectations and will be replaced by the company’s figure in February. Cloud margins rest on an assumed depreciation allocation, and the base case gives TPU hardware the same margin path as services; a hardware-diluted path gives $259.31. Consensus revenue comes from a secondary aggregation and does not split the segments; the split is ours. The Anthropic stake is not disclosed by Alphabet; we use the $128.2bn carrying value, which the 10-Q says consists primarily of one private company; a listing at $1.5tn with a stake near 15% would add about $6.63 a share after tax, at $2tn about $11.37. Waymo’s ≈80% stake is a third-party estimate and its mark dates from February. The fourth-quarter 2026 cash flow is derived from the modelled year less the nine-month actual. Terminal value is 56% of core enterprise value. The reference price is the close two sessions after the results, after the largest one-day fall in the stock’s history; the rating is sensitive to the price chosen.
Alphabet reported its third quarter, the three months to 30 September 2026, after the close on 28 October and filed its Form 10-Q on 29 October. The shares closed at $324.89 on 28 October, fell 12.6% to $283.82 on 29 October, their worst day since listing in 2004, and closed at $279.84 on 30 October.
| Q3 2026 | Value | Prior | Change | Why it matters |
|---|---|---|---|---|
| Revenue | $127.6bn | $102.3bn | +25% | A half-trillion annual run rate; the 13th straight quarter of double-digit growth |
| Search & other | $63.4bn | $56.6bn | +12% | The profit engine; slowest growth in two years |
| YouTube ads | $11.2bn | $10.3bn | +9% | Brand advertising led; living room over a third of ad revenue |
| Google Cloud revenue | $30.7bn | $15.2bn | +103% | Now a quarter of revenue and the main growth source |
| Cloud operating margin | 36.2% | 23.7% | +12.5 pts | Held with TPU hardware in the mix; near-term pressure guided |
| Services operating margin | 40.7% | 38.5% | +2.2 pts | Still rising year on year despite AI serving costs |
| Cloud backlog (vs Q2) | $560.8bn | $514bn | +9% q/q | Just over 55% due within 24 months (just over 50% in Q2) |
| Capital spending | $51.4bn | $24.0bn | +114% | 40% of revenue; the number the valuation depends on |
| Depreciation | $8.4bn | $5.6bn | +49% | The build-out arriving in the income statement |
| Free cash flow | −$9.1bn | $24.5bn | n/m | Twelve-month figure down to $19.7bn |
| Model API tokens a minute (vs Q2) | ≈32bn | ≈23bn | +40% q/q | More than four times a year earlier; supply-constrained |
| Employees | 204,757 | 190,167 | +8% | Headcount growing at a third of revenue’s pace |
Prior is the third quarter of 2025 unless marked. Sources: Q3 2026 release and Form 10-Q; call of 28 October 2026 (tokens); Q2 2026 release (backlog). Margins and changes derived.
One: Cloud doubled, and the margin held. Revenue grew 103% to $30.7bn, up from 82% in the second quarter, with core cloud services, AI infrastructure and AI solutions all contributing and TPU system sales “a meaningful contributor”. Operating margin was 36.2% against 35.6% in the second quarter and 23.7% a year earlier.
Two: Search slowed to 12%. Management named currency as the largest factor, then retail advertisers in Asia hit by the end of the EU’s de minimis exemption and media spending normalising after the World Cup quarter, then strong growth in AI Mode queries that monetise at lower rates. The chief financial officer said the quarter lapped about three points of the late-2025 acceleration and the fourth quarter laps “a couple of points” more, with a somewhat larger currency headwind.
Three: capital spending held for 2026 and rises “significantly” in 2027. The 2026 guide stayed at $195–205bn, which implies $63–73bn in the fourth quarter after $132.0bn in nine months. About 65% of third-quarter spending went to servers and networking. The 2027 figure comes with the fourth-quarter results.
Four: free cash flow was negative for a second quarter. Operating cash flow of $42.3bn, after a $5.2bn payment for the 2018 EU Android fine and a $3.2bn inventory build, did not cover $51.4bn of capital spending. The chief financial officer described TPU systems as a different cash cycle, built only against signed contracts, with cash out before cash in.
Five: the balance sheet is funding the build. Alphabet issued $28.7bn of senior notes and sold 36m shares for $12.4bn through its at-the-market programme in the quarter; long-term debt reached $125.1bn and quarterly interest expense $967m, from $143m a year earlier. No shares were repurchased; $27.6bn of the programme remains.
Every revenue line grew except Network (+1%). United States revenue grew 39%, lifted by Cloud and TPU sales; EMEA grew 9% and Asia-Pacific 12%. Consolidated operating margin rose to 31.8% from 30.5%. Other cost of revenues grew 32% on TPU hardware costs, depreciation, content and infrastructure operations; research and development grew 36% on AI talent and depreciation; general and administrative costs fell 17% on lower legal charges. The Alphabet-level loss widened to $7.7bn from $4.5bn as model development spending rose. Alphabet does not guide revenue or earnings; it reiterated 2026 capital spending and flagged a somewhat larger currency headwind, harder Search comparisons, near-term Cloud margin pressure as rented third-party capacity ramps, continued growth in TPU system sales, and higher depreciation and data-centre costs.
The tone was confident about demand and guarded on numbers. The chief executive, Sundar Pichai, described Gemini 4 as leading at the frontier and the company as still capacity-constrained; the chief financial officer, Anat Ashkenazi, repeated that returns on invested capital should stay “meaningfully above” the cost of capital but declined again to give product-level figures for TPU systems. On Search the message was that usage is strong and monetisation is catching up, with the slowdown framed mostly as currency and comparisons rather than AI substitution.
| Firm | Question | Answer, in short |
|---|---|---|
| JPMorgan, Morgan Stanley | Why is Search revenue slowing while queries grow at a five-year high? | Currency, Asia retail, the World Cup comparison and lower AI Mode monetisation; ads in AI Overviews and AI Mode are scaling |
| Barclays, Bank of America, Goldman Sachs | How much of a gigawatt-scale TPU deal becomes revenue, at what margin, and what if labs build their own sites? | No product-level figures; TPU systems are inventory, not capital spending, and margins are a focus |
| Bank of America, AllianceBernstein | Is agentic search a threat? | “Clearly additive” and an advertising opportunity; consumer roll-out later in 2026 |
| MoffettNathanson | Can you sell TPUs to competitors? | TPUs are a platform that has to win on merit |
| Positive surprises | Negative surprises |
|---|---|
| Cloud revenue about 7% above consensus, according to one sell-side note | Search & other at 12% against about 15% expected; YouTube ads also short |
| Cloud margin held above 36% with TPU hardware in the mix | Operating income about 6–7% below consensus, according to sell-side notes |
| Backlog up about $47bn in a quarter, with more of it due within two years | Free cash flow −$9.1bn, and 2027 capital spending left open-ended |
| Query growth at a five-year high; the Gemini app at one billion users | An $8.7bn loss on equity securities as SpaceX fell, cutting EPS by $0.55 |
Alphabet is a holding company whose value sits almost entirely in Google. It reports three segments and one cost centre. Google Services (75% of third-quarter revenue) is advertising on Search, YouTube and partner sites, plus subscriptions (YouTube Premium and Music, YouTube TV, Google One), app-store fees and devices. Google Cloud (24%) rents computing, storage, data and AI services to companies, sells Workspace and, since the second quarter of 2026, sells complete TPU systems — Google’s own AI chips with networking and software — for installation in customers’ and partners’ data centres. Other Bets (0.3%) holds Waymo and other early ventures. Alphabet-level activities carry the shared cost of building the Gemini models every segment uses, plus corporate costs; they lost $7.7bn in the quarter.
The economic engine is simple to state. Search and YouTube turn attention into advertising revenue at operating margins around 40%, and that cash has historically paid for everything else. Since 2025 the same company has also become one of the largest builders of AI infrastructure in the world, and the build now consumes all of its operating cash. The valuation question is how long that lasts and what the build earns.
| Line | Q3 2026 | Q3 2025 | Change | Share |
|---|---|---|---|---|
| Google Search & otherServices | $63,350m | $56,567m | +12% | 50% |
| YouTube adsServices | $11,207m | $10,261m | +9% | 9% |
| Google NetworkServices | $7,455m | $7,354m | +1% | 6% |
| Subscriptions, platforms & devicesServices | $14,290m | $12,870m | +11% | 11% |
| Google Cloudincl. TPU system sales | $30,709m | $15,157m | +103% | 24% |
| Other Bets | $372m | $344m | +8% | 0.3% |
| Total (incl. hedging) | $127,577m | $102,346m | +25% | 100% |
Source: Q3 2026 results release. Segment operating income: Services $39.2bn (40.7%), Cloud $11.1bn (36.2%), Other Bets −$2.0bn, Alphabet-level −$7.7bn.
AI answers are lifting search usage faster than search revenue. Management says year-over-year query growth is the highest in five years and AI Mode queries run about three times longer than traditional searches, most of them incremental. Revenue has not kept pace: Search & other grew 19%, 17% and 12% in the first three quarters of 2026.
AI compute is still supply-constrained, and every large buyer has committed for years. Microsoft spent $135.9bn on property and equipment over the twelve months to September, Amazon $215.2bn and Meta $104.4bn; Oracle has guided to peak capital spending in its fiscal 2027–2028 and positive free cash flow from the second half of fiscal 2028. Alphabet’s own purchase commitments reached $924.4bn at 30 September, $802.1bn of them multi-year, mostly fulfilled through 2031.
TPUs have become a product. Alphabet now sells TPU systems directly and through Oracle and Crux AI, and says about 100 significant customers use TPUs. Anthropic’s registration statement discloses a commitment to spend at least about $114.5bn with Google through July 2033, plus an estimated $125–150bn of further TPU compute from TPU providers, including Google, for delivery in 2028–2031. The margins of the hardware business are not disclosed.
Antitrust is behavioural, durable and still in the courts. The search judgment of December 2025 is under appeal; the ad-tech court rejected divestiture in September 2026 and ordered conduct remedies, with a final judgment to come; the European Court of Justice ended the Android case in July 2026 and Alphabet paid $5.2bn. Accrued fines and settlements stood at $13.8bn at 30 September.
| Arena | Main rivals | Where Alphabet stands |
|---|---|---|
| Search & answers | OpenAI (ChatGPT), Microsoft (Bing, Copilot), Anthropic (Claude), Perplexity, Meta AI, Apple | Dominant search share; Gemini app at one billion users; query growth at a five-year high |
| Digital advertising | Meta (revenue +26% in the September quarter), Amazon ads, TikTok, Microsoft | The largest ad business; advertising grew 11% in the quarter, slower than Meta |
| Cloud & AI infrastructure | Amazon Web Services, Microsoft Azure, Oracle (revenue +30% in its August quarter), CoreWeave and other neoclouds | Fastest-growing of the three largest clouds at +103%; also sells its own training chips as complete systems |
| Frontier models | OpenAI, Anthropic, Meta, xAI (inside SpaceX), DeepSeek and other open models | Gemini 4 Argon released; Carbon announced at the top of a widely used independent index |
| Video & living room | Netflix, TikTok, Meta, Amazon Prime Video | Living-room viewing over a third of YouTube ad revenue; 150m Music and Premium subscribers including trials |
| Autonomous driving | Tesla, Zoox (Amazon), Baidu Apollo, Pony.ai | Waymo: over 300m fully autonomous miles in 15 US cities; valued at $126bn after its February round |
| Source | GOOGL | MSFT | AMZN | META | Alphabet evidence |
|---|---|---|---|---|---|
| Brand | 5 | 5 | 5 | 4 | “Google” is a verb; the Gemini app at one billion users |
| Network effects | 4 | 3 | 3 | 5 | Searches improve results and attract advertisers; YouTube creators and viewers |
| Switching costs | 3 | 5 | 4 | 2 | Low for consumers; high in Cloud once data, models and agents are built in |
| Cost advantage | 5 | 4 | 4 | 3 | Own TPUs, network and models; TPU systems now sold to third parties |
| Scale | 5 | 5 | 5 | 4 | Among the largest capital budgets in the world; $924bn of purchase commitments |
| Intellectual property | 5 | 4 | 4 | 4 | DeepMind research, Gemini 4, TPU designs (with Broadcom), the Waymo Driver |
| Distribution | 4 | 5 | 4 | 4 | Android, Chrome and defaults, now limited by court-ordered conduct rules |
| Regulation | 2 | 3 | 3 | 2 | Scored as a tailwind: regulation works against Alphabet in the US, EU and UK |
| Data | 5 | 4 | 4 | 5 | Query, video and map data at unmatched breadth; rivals now get parts of the index |
| Customer loyalty | 3 | 4 | 4 | 3 | Habit on the consumer side; the Cloud backlog up 9% in a quarter |
Scores are our judgement. Totals out of 50: Alphabet 41, Microsoft 42, Amazon 40, Meta 36.
Strongest advantages. Vertical integration in AI: Alphabet designs its own training chips at scale, trains a frontier model on them, runs one of the three largest clouds and owns billion-user surfaces to deploy the model into. The third quarter showed the layers reinforcing each other: TPU demand from outside customers, Anthropic among them, lifted Cloud revenue while the margin held, and the same chips serve Gemini inside Search. And monetisable intent at a five-year usage high: people are searching more, not less.
Biggest threats. Answers that do not need an ad: the third quarter is the first in which usage and revenue clearly diverged, and management attributes only part of it to AI Mode. Mandated data sharing gives qualified rivals parts of the index for six years. And customers who are competitors: Anthropic buys TPUs, competes with Gemini and spreads its compute across four chip families.
Direction: stable overall, with the composition shifting faster. The consumer search moat is narrower than five years ago and the third quarter gave the first hard evidence of pressure on its monetisation. The infrastructure moat widened again. Because that moat is capital-intensive, it earns its returns on a much larger asset base than the old one did; that difference is the heart of the valuation.
| Concern | Severity | Evidence and why it matters |
|---|---|---|
| Cash conversion | HIGH | Free cash flow −$5.9bn in Q2 and −$9.1bn in Q3; $19.7bn over twelve months against $73.3bn in 2025; capital spending 37% of nine-month revenue and rising “significantly” in 2027. In our base case free cash flow is negative until 2029 |
| Search monetisation | MED-HIGH | Search & other growth 19% → 17% → 12% while query growth hit a five-year high; management cites lower monetisation of AI Mode queries |
| Circular AI relationships | MED-HIGH | Anthropic has committed at least about $114.5bn to Google; Alphabet may own up to 14.99% of it and has agreed further financing through a 2026 note agreement. Crux AI: $16.8bn in the backlog, while Alphabet backstops up to $10bn of agreements the venture enters into. Backlog concentration is not disclosed |
| Balance-sheet reversal | MED-HIGH | Long-term debt from $46.5bn to $125.1bn in nine months; $42.9bn of common and $19.1bn of mandatory preferred issued in 2026; no buybacks. In our base case Alphabet needs about $130bn more than it generates, after dividends, from the fourth quarter of 2026 to the end of 2028 |
| Fixed commitments | MEDIUM | $924.4bn of purchase commitments, $802.1bn of them multi-year, plus leases not yet commenced: less flexibility than the word “guidance” suggests |
| Earnings quality | MEDIUM | $127.3bn of nine-month gains on equity securities in net income; the third quarter reversed with an $8.7bn loss. GAAP EPS and P/E say little about earnings power |
| Legal and regulatory | MEDIUM | $13.8bn of accrued fines and settlements (deducted as debt here); search and ad-tech remedies under appeal; a €3.0bn EU ad-tech fine under appeal |
| Depreciation lives | MEDIUM | Servers and network equipment over six years while TPU generations turn faster; shorter lives would cut reported margins, not cash |
| Inventory and hardware | MEDIUM | Inventory from $2.4bn to $13.3bn in nine months; TPU margins not disclosed |
| Stock-based compensation | MEDIUM | $8.6bn in the quarter (+36%); equity sold to fund award taxes; a real cost no longer offset by buybacks |
| Governance | LOW | Class B shares with ten votes give the founders control; outside holders cannot force a change in capital allocation |
| Insider dealing | LOW | Forms 4 since July: tax withholding on vesting, gifts and small planned sales, the largest about $3.1m; no open-market purchases |
Severity is our judgement. Sources: Q3 2026 release and Form 10-Q, notes 2, 3, 6, 7 and 10; Anthropic Form S-1, 26 October 2026; Forms 4, July–October 2026.
Overall red-flag score: 6 out of 10. The concerns cluster around cash: how long capital spending absorbs all of operating cash flow, how much of the backlog sits with counterparties Alphabet finances, and whether Search revenue can keep up with Search usage as answers replace links. Alphabet held $174.7bn of cash and debt securities at 30 September against $126.1bn of debt, so the funding need is a choice about leverage, not a solvency question.
| Strength | Weight | Evidence |
|---|---|---|
| Contracted Cloud demand | HIGH | $560.8bn of backlog, up $47bn in a quarter, just over 55% due within 24 months; Anthropic’s take-or-pay commitment of at least about $114.5bn; about 100 significant TPU customers |
| Cloud profitability | HIGH | Operating margin 36.2% with TPU hardware in the mix, from 23.7% a year earlier; operating income tripled |
| Full-stack cost position | HIGH | Own chips, models, network and distribution; TPUs used by rival model developers, Anthropic among them |
| Search usage | HIGH | Query growth at a five-year high; AI Mode queries three times longer and mostly incremental; ads scaling in AI Overviews and AI Mode |
| Stakes below market evidence | MEDIUM | Private stakes carried at $128.2bn against a $48.6bn cost; a listing at $1.5–2tn with a stake near 15% would be worth $225–300bn before tax (+$6.63 to +$11.37 a share after tax) |
| Cash-tax deferral | MEDIUM | Full first-year expensing of US equipment: about $95bn of deferral over 2026–2030 in our base case |
| A cheaper starting point | MEDIUM | 18.6 times 2027 consensus earnings against 26.1 for Microsoft and 22.5 for Amazon |
The balance of the evidence. Overall green-flag score: 7 out of 10. The business evidence remains stronger than the cash evidence, and at this price the market does not assume the best of it. Both sides are carried into the valuation once: backlog conversion and Cloud margins in the base case; faster Search monetisation and lower capital spending in the bull case, with the same weight as the bear case’s disruption; the stakes at their latest marks, with listing values shown as a sensitivity.
| Dimension | Score / 5 | Assessment |
|---|---|---|
| CEO track record | 4 | Sundar Pichai: revenue from about $75bn in 2015 to a $500bn run rate; Cloud growing 103%; Gemini 4 at the frontier; liability findings in two US antitrust cases, but no breakup |
| CFO credibility | 4 | Anat Ashkenazi: a clear funding hierarchy and candour that free cash flow stays under pressure; $28.7bn of notes and $12.4bn of share sales executed in one quarter |
| Guidance accuracy | 3 | The 2026 capital-spending guide was raised twice after February, then held this quarter; no revenue or margin targets to hold management to |
| Transparency | 2 | Discloses backlog, purchase commitments and token volumes; no TPU revenue or margin, no segment depreciation, no 2027 capital figure; the size of the Anthropic commitment came from Anthropic’s filing, not Alphabet’s |
| Capital allocation | 3 | Bold and possibly right; also a sharp reversal from $62bn of buybacks in 2024 to issuing equity and nearly tripling debt |
| Acquisitions | 3 | Wiz (about $32bn) and Intersect ($5.9bn, renewable energy) fit the strategy; returns still to be shown |
| Buybacks | 2 | None in 2026, against $39.8bn in the first nine months of 2025 |
| Dilution | 3 | 36m shares sold through the programme in the quarter; the mandatory preferred converts in 2029 |
| Insider ownership | 4 | The founders’ Class B stakes: strong alignment on long-term value, weak accountability |
| Compensation | 3 | Stock pay of $8.6bn in the quarter, about 7% of revenue |
| Board quality | 3 | Experienced directors; founder voting control limits independence |
| Communication | 4 | Detailed calls and consistent operating metrics; prefers qualitative answers on returns |
Composite about 3.2 out of 5. Sources: Q3 2026 release, Form 10-Q and call; 2026 proxy statement; Forms 4.
Do they act like long-term owners? Yes, in the sense that matters most to them: they are taking several years of weak free cash flow for a position in AI infrastructure they believe will compound, and the founders’ control makes that possible. The cost is accountability. Management has not yet given outsiders the numbers needed to check the return on the largest investment programme in its history, and the most important single figure about its largest Cloud customer reached investors through that customer’s prospectus.
| Window | Event | Upside | Downside | Date · source |
|---|---|---|---|---|
| November 2026 | Anthropic IPO pricing | A valuation well above the $128.2bn carrying value; Alphabet’s option to buy up to $10bn alongside | A valuation below the press figures | Reported · press, S-1 |
| Mid-November | Q3 13F filings | — | Sales of SpaceX shares | Firm · filing calendar |
| Nov–Dec 2026 | Ad-tech final judgment and appeal notices | Narrow wording | Broader conduct rules | Likely · 10-Q |
| Q4 2026 | Gemini 4 Carbon wider release; agentic features for consumers | Frontier position at lower serving cost | A slower roll-out | Likely · company |
| Late 2026 – early 2027 | Search case argued on appeal | Data-sharing remedies narrowed | Remedies widened on the cross-appeal | Reported · press |
| Early Feb 2027 | Q4 results and 2027 capital-spending guide | 2027 at or below about $300bn with Cloud growth intact | Well above $300bn without more revenue | Firm · reporting pattern |
| H1 2027 | Search comparisons ease | Growth re-accelerates | Growth keeps slowing | Likely · company commentary |
| Late Apr 2027 | Q1 2027 results; Google Cloud Next | TPU revenue ramp with the margin held | Hardware diluting the Cloud margin | Firm · reporting pattern |
| Within 12 months | Short-term SpaceX sale restrictions lapse ($75.2bn restricted at 30 September) | A source of funding for the build | — | Firm · 10-Q |
| Late Jul 2027 | Q2 2027 results | Free cash flow turning up | Capital spending rising again | Firm · reporting pattern |
| Ongoing | At-the-market programme ($27.6bn left); dividend of $0.22 a quarter | Buybacks resuming | Faster share sales | Firm · company |
“Reporting pattern” means the date follows the company’s usual calendar; “reported” means press reports, not company announcements. The single most informative items are the 2027 capital-spending guide and Search growth once the late-2025 surge is lapped, because the base case already assumes consensus revenue.
BULLGrowth. The market sold the best cloud quarter any hyperscaler has reported because Search grew 12% instead of 15%. Cloud revenue doubled to $30.7bn, the backlog grew $47bn in three months to $561bn, and more of it now arrives within two years. Anthropic alone has committed at least about $114.5bn; about 100 significant customers use TPUs; Oracle and Blackstone are distributing them.
BEARUsage at a five-year high and Search growth at a two-year low is what monetisation leakage looks like, and management conceded that AI Mode monetises at lower rates. Currency cost about one point of total growth in the quarter; Search slowed by five. The segment that pays for everything else is decelerating at exactly the moment the bill arrives.
BULLValuation. At about 18.6 times next year’s earnings the shares are cheaper than Microsoft and Amazon for faster growth. The private stakes are carried at $128bn while the Anthropic listing is discussed at valuations that would put Alphabet’s share at $225–300bn. The market is paying for the base case and getting the option on the rest for free.
BEAREarnings do not charge the capital spending; our base case has it exceed depreciation by $1.06tn in five years. On free cash flow the shares are at 175 times, and even the base case leaves almost nothing for shareholders before 2029.
BULLBusiness quality. Cloud’s margin held above 36% with hardware in the mix. Alphabet is the only company that designs its own training chips at scale, trains a frontier model on them, runs a top-three cloud and owns billion-user surfaces.
BEARHardware sales inflate Cloud revenue at margins nobody will disclose. And the largest identified customer is also a competitor that spreads its compute across four chip families.
BULLRisks and financials. $174.7bn of cash and debt securities; a funding hierarchy of cash, then debt, then equity; returns on capital “meaningfully above” the cost of capital, management says.
BEARFree cash flow negative for a second quarter, $924bn of purchase commitments, $126bn of debt and stock sold through an at-the-market programme. In our base case Alphabet needs about $130bn more than it generates by the end of 2028.
BULLManagement and catalysts. A team that took Cloud from a laggard to the fastest-growing major cloud. The Anthropic listing and easier Search comparisons both arrive within months.
BEARThe growth is partly circular: Alphabet owns part of Anthropic, has agreed to keep financing it and counts its commitments in the backlog; it backstops up to $10bn of Crux AI’s agreements and books a $16.8bn contract with it. And the February capital-spending guide can move the value more than any of those catalysts.
The bull has the better of the argument about Cloud; the bear has the better of the argument about cash. Cloud’s evidence is recent, disclosed and contracted. The capital-spending path and its returns are the least disclosed numbers in the company, and they decide the value. On Search the evidence is genuinely mixed — one quarter of divergence, with plausible non-structural explanations — and it will take two or three more quarters to settle. At $279.84 the shares sit on our base case, between the bear at $105.83 and the bull at $426.00: fairly valued. What would change it: 2027 capital spending well away from $300bn without a matching revenue change; Search re-accelerating or slipping further as comparisons ease; a disclosed TPU margin; free cash flow turning up before 2029. Verify next: the Anthropic listing price and Alphabet’s resulting stake, the fourth-quarter capital-spending guide, and any disclosure of backlog concentration.
Alphabet owns Google. It makes most of its money from ads next to search results and on YouTube, and a fast-growing share from renting computing power and AI to other companies, and now from selling its own AI chips. It also owns Waymo, the robotaxi company, and stakes in SpaceX and Anthropic.
Advertisers pay when people search or watch videos; businesses pay to use Google’s data centres, AI models and chips; consumers pay for YouTube Premium and Google One. Out of every $100 of sales last quarter, about $32 was operating profit.
Its cloud business doubled in a year, and companies have signed up for $561 billion of future cloud spending. At the same time, the search business, which pays for everything, grew more slowly than people expected.
Alphabet is spending around $200 billion this year, and more next year, on data centres. Whether that spending earns more than it costs is the whole question, and the company does not yet publish the numbers that would let anyone check.
On our most likely assumptions the shares are worth about $245 to $320 each, most likely $277.04. At $279.84 we call them fairly valued: the price matches what the most likely path is worth. The shares fell 12.6% in one day after these results, and they could keep moving sharply either way as the spending plans and the Anthropic listing become clearer.
Right: ads inside AI answers catch up with usage; cloud and chip sales keep surprising; spending levels off sooner; Anthropic lists at a high value.
Wrong: AI chat answers take more of Search’s ad money; the data centres earn less than they cost; the company keeps needing to borrow or sell shares; the values of SpaceX and Anthropic fall.
The Anthropic listing, expected in November. Alphabet’s spending plan for 2027, in early February. And whether Search growth picks up again in the first half of 2027.
| Status | What it covers |
|---|---|
| Verified from Alphabet filings and releases | Revenue by line and segment, segment operating income, cash flow, capital spending, depreciation, stock compensation, the balance sheet, debt and the third-quarter note issues, the at-the-market programme, the mandatory preferred, the share count, revenue backlog and Crux AI, fair values and costs of equity securities, SpaceX restrictions, accrued fines and settlements, purchase commitments, the Waymo term loan, legal matters and Forms 4; annual figures for 2023–2025 from the Forms 10-K |
| Verified from other filings and market data | Anthropic’s Form S-1 (26 October 2026): its commitment to Google, further TPU compute, the 2026 note agreement and ownership cap, and the concurrent placement option; peers’ Forms 10-Q and Oracle’s Investor Day release; closing prices on 28, 29 and 30 October 2026 and SPCX on 30 September; the 10-year Treasury yield (Federal Reserve H.15) |
| Third-party, attributed | Call statements from published transcripts and summaries (TradingKey, GuruFocus via Yahoo Finance), checked against the release and Form 10-Q; consensus revenue and EPS from a secondary aggregation; the Search expectation, operating-income shortfall and Cloud beat from sell-side notes reported by Investing.com; 2027 capital-spending expectations and the 29 October fall from Fortune; a sell-side Search multiple from Fortune’s report of a Bank of America note; the Anthropic listing timetable and valuation range from press reports |
| Our estimate or judgement | The segment split of consensus revenue; revenue after 2028; every margin, Alphabet-level, capital-spending and inventory path; the depreciation allocation and asset lives; the fourth-quarter 2026 stub; Waymo’s stake and cost; the cost basis of other listed holdings; the scenario weights; the moat, flag and management scores |
| Still open | 2027 capital spending; TPU revenue and margins; how much of the backlog is Anthropic’s; Alphabet’s Anthropic stake and the listing price; whether any SpaceX shares were sold after 30 September |
Method. Unlevered discounted cash flow in sum-of-the-parts form, methodology version 1.3: core operations on an explicit path to 2040, the fourth quarter of 2026 as a stub and each later year at its mid-point, the Gordon terminal value at 3.0% discounted from the end of 2040; the SpaceX shares, other listed holdings, private-company stakes and Waymo at their latest marks after tax; everything valued at 31 October 2026 from the 30 September 2026 balance sheet. One discount rate (9.5%) and one terminal growth rate are used across all scenarios, and each material risk is priced once. Stock-based compensation is a cost inside the margins; the mandatory preferred is counted as converted at the modelled price; accrued fines and settlements are deducted as debt. Scenario weights are our judgement, the fair-value band is the corners of the WACC 9–10% by terminal growth 2.5–3.5% grid, and the reverse valuation is run against the market price. The rating changes to overvalued if the price rises above $320.06 with no change in the evidence, and to undervalued below $245.08; the conclusion is most sensitive to steady-state capital spending, Cloud growth and the discount rate, in that order. Ratings are undervalued, fairly valued or overvalued according to where the price sits relative to the band; Sonde does not issue buy, sell or hold recommendations or price targets. The reference price used throughout is $279.84, the Nasdaq close on 30 October 2026, and every multiple in this report scales directly with it. Ratings are reviewed each quarter on results and immediately on a material contract, a financing event or a share-price move of more than 25%.