Alphabet GOOG

Reference table

1. What the business actually is

The 10-Q filed 2026-07-23 states the segmentation plainly: "We report our segment results as Google Services, Google Cloud, and Other Bets", with certain costs held outside the segments because "they represent Alphabet-level activities" — a line the filing notes "primarily reflect expenses related to our shared AI research and development." That Alphabet-level bucket is new relative to the older three-segment presentation and is worth naming: centralised advanced research is now carried above the segments rather than inside them.

Revenue by type, FY2025 (from seg_prod, with one vendor label corrected against the filing):

Vendor label defect, corrected. The vendor's FY2025 row labels the $48.030B line "Google Inc." — an opaque, meaningless label. The identical line in the FY2024 row is labelled "Google Subscriptions, Platforms, And Devices", and the 10-Q uses exactly that category name ("other sources ('Google subscriptions, platforms, and devices')"). We use the filing's label. The vendor also carries a spurious "Other Segments" row of −$127M in FY2025: the six real lines sum to precisely $402.963B, the reported total, so the −$127M is a double-count and is discarded.

Revenue by geography, FY2025: United States $194.229B (48.2%), EMEA $117.152B (29.1%), Asia-Pacific $67.680B (16.8%), Americas excluding the US $23.902B (5.9%). The sum is exactly $402.963B — this table is clean. Note that a business with 51.8% of revenue outside the United States carries meaningful currency and regulatory exposure, and the European antitrust matter discussed in Section 8 is a direct consequence.

Two acquisitions closed in Q1'26 that the vendor's annual data cannot see. The 10-Q discloses that on 2026-03-11 Alphabet completed the acquisition of Wiz for $29.5B after purchase-price adjustments, recorded in the Google Cloud segment; and on 2026-03-10 it completed the acquisition of Intersect, a developer of renewable energy, for $5.9B, described as enabling "acceleration of data center capacity and energy development." The vendor's most recent cash-flow year (FY2025) shows acquisitions of $1.592B. The vendor is a full year behind on $35.4B of deal activity, and both deals bear directly on the cloud and power-capacity thesis. A pending divestiture of the GFiber interest is also disclosed, expected to yield $1.5B in cash and a $2.0B interest, with GFiber's results remaining in Other Bets until close.

Headcount. The 10-Q states 198,933 employees as of 2026-06-30. The vendor profile says 190,820. The filing wins; we use 198,933.

2. The earnings illusion — the single most important thing on this page

Alphabet's reported GAAP earnings for the last two quarters are dominated by unrealised marks on private equity stakes and are close to useless as a measure of the business.

The 10-Q filed 2026-07-23 is unambiguous:

> "OI&E of $98.0 billion for the three months ended June 30, 2026 included net gains on equity securities of $99.0 billion, primarily related to unrealized gains in our equity securities portfolio from SpaceX and a private company."

and, in the segment discussion:

> "OI&E, net increased $95.3 billion and $121.9 billion from the three and six months ended June 30, 2025 to the three and six months ended June 30, 2026, respectively, primarily related to unrealized gains in our equity securities portfolio."

The consequence is visible in the quarterly series:

In Q2'26, reported net income exceeded operating income by $71.4B. Revenue for the quarter was $119.796B and reported profit was $112.193B — a 93.7% net margin, which is not an economic fact.

The decisive corroboration comes from the estimate block itself. Consensus for the next quarter (2026-10-28) is EPS of $3.02. Consensus for full-year FY2026 is $20.259 on 32 analysts. Reconcile them:

> Q1'26 actual $5.11 + Q2'26 actual $9.11 + Q3'26E $3.02 + Q4'26E ≈ $3.02 = $20.26

That is the FY2026 consensus, to the cent. The analysts are carrying the reported first-half figures — paper gains and all — and modelling roughly $3.02 a quarter for the clean second half. The market's own view of Alphabet's underlying quarterly earnings power is therefore approximately $3.02, i.e. ~$12.1 annualised, not $19.91 trailing.

The multiple ladder, stated honestly:

Every screen in the market shows this company at 18.9x. It is at 25.0x on the first clean forward year. That is a 32% difference in the headline valuation and it is the reason this dive exists.

The same distortion runs through the trailing ratio block and we reject it wholesale. ratios_ttm.ebitMarginTTM reports 67.4% and ebitdaMarginTTM 73.0% against an operatingProfitMarginTTM of 33.1% — the vendor's "EBIT" is scooping the equity gains into the operating line. pretaxProfitMarginTTM of 67.1% and returnOnEquityTTM of 50.8% are contaminated the same way. ebtPerEbitTTM of 2.026 is the vendor unknowingly documenting its own problem: pre-tax income is twice operating income. All of these are rejected (Section 8).

One legitimate operating charge in the quarter, for completeness. The 10-Q discloses a $2.1B legal accrual from a Stockholm Patent and Market Court decision in a private action, of which $1.5B sits in general and administrative expenses within Google Services and $581M in other income. That charge is in the $40.770B of operating income — so Q2'26 operating income is understated by $1.5B relative to a clean run-rate, and the true underlying operating margin was closer to 35.3%.

3. Growth and margins — where the acceleration actually is

Annual history:

Revenue growth has accelerated for three consecutive years and then accelerated again inside 2026: +8.7% (FY2023), +13.9% (FY2024), +15.1% (FY2025), +21.8% (Q1'26), +24.2% (Q2'26). Gross margin has expanded 630 basis points since FY2022 and reached 61.6% in Q2'26 — the highest in the series.

The 10-Q supplies the drivers, which the vendor cannot:

• "Revenues were $119.8 billion, an increase of 24% year over year, primarily driven by an increase in Google Services revenues of $12.0 billion, or 15%, and an increase in Google Cloud revenues of $11.1 billion, or 82%." Cloud contributed 48% of the absolute revenue growth from 15% of the base.
• "Google Cloud revenues increased $11.1 billion ... primarily driven by growth in Google Cloud Platform largely from infrastructure and platform services. In addition, in the second quarter of 2026, we began recognizing revenue from the sale of TPU systems." A new hardware revenue line, first recognised in this quarter. This is the most consequential single sentence in the filing: the company has begun selling the custom silicon it previously only consumed internally.
• "Google Cloud operating income increased $6.0 billion and $10.4 billion from the three and six months ended June 30, 2025" — the segment's profitability is scaling faster than its revenue.
• "Google Services operating income increased $6.5 billion and $14.4 billion."
• Traffic acquisition cost rate fell from 20.6% to 19.8%, "primarily due to a revenue mix shift from Google Network properties to Google Search & other properties." Google Network revenue is the one line in decline (−1.9% in FY2025, and "decreased $51 million and $336 million" in Q2 and H1 respectively), and its shrinkage is margin-accretive.
• Cost of revenues rose $6.9B in the quarter, "due to an increase in other cost of revenues and TAC of $5.4 billion and $1.5 billion", with the other-cost increase driven by depreciation, inventory, and content acquisition.
• R&D rose $4.4B in the quarter, of which $592M was depreciation.
• Other Bets operating loss increased $553M in the quarter and $1.4B in the half. The loss is widening, not narrowing.

The margin warning is embedded in those last three bullets: depreciation now appears as a growth driver of cost in cost of revenues, in R&D, and implicitly everywhere else. Consensus reflects it — modelled EPS grows ~17.6% annually from FY2027 to FY2030 against ~18% revenue growth, i.e. no net margin expansion at all over four years, despite the gross-margin trend of the last four.

Consensus revenue path (analyst counts in brackets): FY2026 $495.176B (29), FY2027 $605.263B (31), FY2028 $720.255B (14), FY2029 $815.730B (26), FY2030 $936.258B (15). That is +22.9%, +22.2%, +19.0%, +13.3%, +14.8%. Note the thin FY2028 and FY2030 coverage.

4. Balance sheet — and why every screen has the wrong numbers

The vendor's most recent balance sheet is FY2025 (2025-12-31). The 10-Q filed 2026-07-23 is two quarters newer and contradicts it in both directions. The filing wins.

What we use. The 10-Q states: "As of June 30, 2026, we had $242.5 billion in cash, cash equivalents, and short-term marketable securities" and "As of June 30, 2026, we had long-term debt outstanding with a total carrying value of $98.2 billion." We use those two figures. They imply roughly $144B of net cash before finance leases, against the vendor's reported net debt of $28.583B — a swing of over $170B on a $4.6T company.

Three separate vendor defects stack here.

1. Staleness. The vendor's newest balance sheet is 2025-12-31; two quarters of enormous capital-structure activity have happened since.

2. Short-term investments ignored in netDebt. Even on the vendor's own FY2025 sheet, netDebt of $28.583B is total debt ($59.291B) less cash and cash equivalents only ($30.708B), silently discarding $96.135B of short-term investments. Correctly computed on the vendor's own data, FY2025 was net CASH of $67.552B, not net debt of $28.583B.

3. Enterprise value. km_ttm.enterpriseValueTTM reports $4,627.207B. Rebuilt from the vendor's own inputs (market cap $4,555.211B + debt $59.291B − cash and short-term investments $126.843B) it is $4,487.659B, a 3.0% residual. Rebuilt from the filing's figures it is materially lower still. All EV-derived multiples — evToSalesTTM 10.38x, evToEBITDATTM 14.21x, netDebtToEBITDATTM 0.175x, enterpriseValueMultipleTTM 14.21x — are rejected, both because the inputs are wrong and because the EBITDA denominator is contaminated per Section 2.

The capital-structure reversal, from the filing, in one place. This is the single most important non-obvious fact about Alphabet today:

> "In the three and six months ended June 30, 2026, there were no repurchases of the company's Class A or Class C shares." — 10-Q, 2026-07-23. Of the $70.0B programme authorised in April 2025, $69.5B remained available at 2026-06-30.

Compare that with the recent past: $45.709B of buybacks in FY2025, $62.222B in FY2024, $61.504B in FY2023, $59.296B in FY2022. Alphabet repurchased roughly $229B of its own stock across four years and has now stopped completely while selling equity at $351-355 a share. Whatever else it signals, it is not a management team that regards the shares as cheap at these levels — a fact worth setting directly against the price at $375.35.

The mandatory convertible preferred deserves its own note. It carries a 6.25% coupon, pays $12.15 per share quarterly (≈$0.60 per depositary share), converts automatically on or about 2029-05-15, and holders have no voting rights with limited exceptions. Annual preferred dividends run roughly $924M. Capped call transactions were purchased to limit the conversion dilution. Net income available to common shareholders is now calculated after deducting accrued preferred dividends — a small but permanent new drag that did not exist a year ago, and one that no trailing per-share figure in the vendor file reflects.

Leases — the largest number nobody quotes. From the 10-Q:

> "As of June 30, 2026, we have entered into leases, primarily related to data centers, that have not yet commenced with future lease payments of $85.2 billion. These leases will commence between 2026 and 2031 with non-cancelable lease terms between one and 26 years."

Plus a $5.8B short-term lease commitment entered in June 2026 commencing in Q3'26, and $21.3B of undiscounted operating-lease and $2.9B of finance-lease payments already on the books. $85.2B of not-yet-commenced, largely non-cancellable data-centre lease obligations does not appear in any leverage ratio in this file, and is larger than Alphabet's entire reported debt load twelve months ago.

Other balance-sheet facts, FY2025: total assets $595.281B; property, plant and equipment $261.818B (up from $184.624B in FY2024 — a $77.2B increase in one year); goodwill $33.380B; long-term investments $68.687B; total equity $415.265B; retained earnings $324.055B. Book value per share $52.71, so the stock trades at 7.12x book. Debt to equity 0.176; financial leverage 1.44x; interest coverage 296x.

5. Cash flow — the year free cash flow turned negative

The 10-Q states it directly: "During the six months ended June 30, 2025 and 2026, we spent $39.6 billion and $80.6 billion on capital expenditures, respectively. In 2026, we expect to significantly increase, relative to 2025, our investment in our technical infrastructure." Capital spending has more than doubled year over year in a single half, and management says it will rise further. Depreciation on property and equipment rose from $9.5B to $13.6B across the same halves — meaning the depreciation is arriving with a lag, and the reported-margin cost of the 2026 programme is mostly still ahead.

With Q2'26 operating cash flow of $39.1B against capex of $44.9B, Alphabet's free cash flow was negative $5.8B in the quarter. The vendor's trailing metrics already register the effect without explaining it: freeCashFlowYieldTTM of 1.17% and priceToFreeCashFlowRatioTTM of 85.8x. Those numbers are arithmetically correct and, for once, we accept them — but only with the causal explanation attached.

Distributions. The common dividend was raised in April 2026 to $0.22 per share quarterly from $0.21, a 5% increase; that is $0.88 annualised, a 0.23% yield at $375.35. Total common dividends were $1.3B (Class A), $184M (Class B) and $1.2B (Class C) in Q2'26 — roughly $2.7B a quarter. Preferred dividends add ~$231M a quarter from August 2026. Total shareholder yield is now approximately 0.23% and falling, against roughly 1.0-1.5% a year ago once buybacks are counted.

6. Valuation — priced in or room?

At $375.35 (market cap $4,555.2B):

Peer context. The vendor peer set is a reasonable mega-cap technology cohort: Apple ($4,544.0B), NVIDIA ($5,133.4B), Microsoft ($3,659.4B), Broadcom ($1,989.4B), Taiwan Semiconductor ($2,163.6B), Meta ($1,497.8B), Cisco ($479.8B), plus three that do not belong in the comparison (ExlService at $5.2B, Flex at $46.4B, Garmin at $59.3B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The observation that stands on its own: at $4.555T, Alphabet is now the second-largest company in its own peer list.

The vendor's scoring model rates Alphabet B+ with an overall score of 3 — return-on-equity and return-on-assets sub-scores of 5, but price-to-book 1 and price-to-earnings 2. Note that the ROE sub-score is built on the contaminated 50.8% trailing figure and should be discounted accordingly.

6a. What today's price assumes (the inversion)

At $375.35 — 25.0x FY2027 consensus EPS of $14.994 — the price embeds roughly the following falsifiable claims:

• Revenue compounds ≥19% annually through FY2028, to $720.255B. (Consensus-derived: $495.176B FY2026E on 29 analysts, $605.263B FY2027E on 31, $720.255B FY2028E on only 14.) This is the assumption the recent prints most support — Q2'26 grew 24.2% — and the one with the thinnest analyst coverage at the far end.
• Clean EPS reaches ~$17.68 by FY2028 and ~$24.37 by FY2030, i.e. clean earnings roughly double from the current ~$12.1 annualised run-rate in four years. (Consensus-derived; 27 and 16 analysts respectively.)
• The market keeps paying ~25x forward clean earnings, i.e. no de-rating despite free cash flow having turned negative, despite roughly $50B of equity and preferred issued in a single quarter, and despite a buyback switched off with $69.5B of authorisation unused. (Our reading of the current multiple against the FY2027 consensus.)
• The $80.6B half-year capital programme earns a return above the cost of the capital raised to fund it — including 6.25% on the preferred. (Our number, derived from the 10-Q capital-raise disclosures.) Return on invested capital is currently 15.1%; the programme must at least sustain that.
• Advertising economics survive. (Our reading of the risk, not a consensus estimate.) Google Search & other is 55.7% of revenue and is growing 13.4%; the price assumes no structural break in how that revenue is generated. This is the most fragile assumption in the price, and it is the one our knowledge base most directly disputes — one tracked source argues that "ad-based revenue loses value once agents, not humans, make purchase decisions." Nothing in the current numbers tests it, which is exactly why it is fragile: it is not falsifiable on a quarterly cadence.

6b. The return bridge (why the multiple moves)

Expected return over 12-24 months decomposes as: clean EPS growth (+18-24% from the current run-rate toward the FY2027 consensus of $14.99) + multiple drift (we assume modest COMPRESSION, from 25.0x to ~23-24x forward) + shareholder yield (~0.23%, dividend only — there is no buyback).

Our base case assumes mild multiple compression, and we should say why rather than dress it as conservatism. Three reasons. First, the mechanical one: 25.0x FY2027E becomes 21.2x FY2028E at a constant price purely by rolldown, and rolldown is not re-rating — but the level the market pays should fall as the earnings base matures and as depreciation from the capital programme compresses incremental margins. Second, the capital structure has changed: a company issuing equity and 6.25% preferred to fund capex is a different security from one retiring 1.5% of its share count a year, and equity issuance at $351-355 is a management signal about value that argues against multiple expansion at $375. Third, free cash flow conversion has collapsed — 1.17% free cash flow yield is a genuinely poor number for a mega-cap, and multiples on companies with negative free cash flow are historically less stable than multiples on companies without.

Almost all of the expected return in the base case is therefore earnings growth, not re-rating. Base fair value of $402 is 26.8x FY2027E — a whisker of expansion, essentially a hold of the multiple — reached by earnings compounding into it. That produces roughly +7.1% plus a 0.23% yield. If the multiple compresses to 22x FY2027E the stock is worth $330, below spot. The thin margin of safety is the verdict.

The bull case at $510 requires 34x FY2027E, roughly the street's high target of $515. That leg needs the market to pay a growth multiple for a company whose consensus models no margin expansion for four years, and it is the fragile leg.

6c. Variant perception (where we differ, what would surprise)

• We differ from the entire screening universe on what Alphabet's multiple actually is. Every trailing-earnings screen shows 18.9x. The honest figure, on the first uncontaminated consensus year, is 25.0x — and the current clean run-rate implies 31.1x. The reconciliation is exact and reproducible: FY2026 consensus of $20.259 equals the two reported quarters ($5.11 + $9.11) plus roughly $3.02 for each of the remaining two. Anyone buying Alphabet because "it's only 19x" is buying a different company from the one that exists. Watchable number: the Q3'26 print on 2026-10-28 against the $3.02 consensus — that is the first clean quarter, and it will make the real earnings power visible to everyone.
• We differ on the balance sheet, in the company's favour and then against it. Screens show net debt of $28.583B; the vendor's own FY2025 data corrected for short-term investments gives net cash of $67.552B; and the 10-Q gives $242.5B of cash and short-term marketable securities against $98.2B of long-term debt as of 2026-06-30. Alphabet is far more liquid than the screens show. But the same filing shows $85.2B of not-yet-commenced lease obligations that appear in no leverage ratio anywhere, plus $51.8B of new notes in six months. The net risk position is roughly unchanged; both halves of it are mis-stated.
• We differ on the significance of the capital reversal, which we think is under-discussed. A company that repurchased $229B of stock over four years bought nothing in the first half of 2026, sold $30.5B of common equity at $351-355, issued $19.0B of 6.25% preferred, and authorised a further $40.0B at-the-market programme it has not yet touched. Our knowledge base contains a bullish reading of this (that spending is forced by demand and the return will come) and a bearish one (that each new dollar of capital buys less compute as input prices rise). We think the correct reading is simply that the risk profile has changed and the multiple has not. Watchable event: any drawdown on the $40.0B ATM programme, or a resumption of buybacks.
• We do NOT differ on the growth. Revenue accelerating from +15.1% to +24.2%, cloud at +82%, a new accelerator-systems revenue line, gross margin at a series high of 61.6% — this is genuinely excellent and we make no attempt to argue otherwise. Our disagreement with the bulls is about price, not about the business.
• Positive surprise that would force a re-rate: Google Cloud operating margin expanding while the 2026 depreciation flows through — that would prove the capital programme is investment rather than expense. Alternatively, a resumption of buybacks against the $69.5B authorisation, or accelerator-system revenue disclosed as a separate line of material size. Watchable numbers: Google Cloud segment operating income (rose $6.0B year over year in Q2'26) and the capex-to-operating-cash-flow ratio (0.713 trailing, but 1.15 in Q2'26 alone).
• Negative surprise that would break the thesis: an advertising revenue deceleration in Google Search & other below ~10% year over year, which would simultaneously validate the agent-disintermediation bear case and remove the cash flow that funds the capital programme. Secondarily: an adverse antitrust remedy affecting distribution agreements, or a downward revaluation of the equity stakes that would reverse the Q2'26 gain as violently as it arrived. Watchable numbers: Google Search & other growth (currently +13.4% FY2025, and "increased $9.1 billion" in Q2'26); the other-income line at each print.

Synthos fair values

All three anchors are multiples of FY2027 consensus EPS of $14.994 (34 analysts — the deepest coverage in the estimate block), cross-checked against FY2028E of $17.679.

• Bear ~$270 — 18.0x FY2027E. Cross-check: 15.3x FY2028E. The scenario: the capital programme's depreciation compresses margins faster than revenue grows, free cash flow stays negative into 2027, the at-the-market programme is drawn, and the market re-rates a capital-consuming business toward a market multiple. Note that $270 is still above the 200-DMA of $326.76 only in the sense that it would break it — this is a genuine de-rating scenario, not a crash. −28.1%.
• Base ~$402 — 26.8x FY2027E. Cross-check: 22.7x FY2028E; 33.3x the current clean run-rate. The scenario: revenue compounds as consensus models, cloud keeps taking share, the multiple broadly holds, and earnings growth does the work. +7.1%, plus a 0.23% dividend.
• Bull ~$510 — 34.0x FY2027E. Cross-check: 28.8x FY2028E; essentially the street's high target of $515. The scenario: the accelerator-systems business becomes a material third-party franchise, cloud margins expand through the depreciation, and the market pays a genuine growth multiple. +35.9%.

Base is 7.1% above spot; asymmetry is roughly 1.3:1 to the upside (28% down, 36% up) — the thinnest risk-reward in this batch relative to business quality. Excellent company, adequate price. Hold.

7. Knowledge base — 402 claims, and the most divided lane we track

Tracked knowledge-base claims naming Alphabet: 510 raw hits, of which 402 are entity-level (the company appears in the structured entities field rather than merely in the claim text). Search covered GOOG/GOOGL (case-sensitive), Alphabet, Google, the chief executive's name, and four product-level terms covering the research unit, the assistant product, the autonomous-transport unit and the video platform, across all 51,928 knowledge-base records.

Discards and weighting, stated openly. 108 hits were thesis- or category-level only and are excluded from the entity count. Ten claims tagged to a quarantined-misattribution source are excluded outright under the standing attribution policy. Management-lane claims — 25 from the chief-executive lane, 29 from the research-unit lane, 7 tagged to the investor-relations lane — carry skill 0.5 and are half-weighted per house rule; where we cite one we say so. Of the product-term matches, 22 relate to the autonomous-transport unit, 12 to the video platform and 7 each to the research unit and the assistant product; these are treated as company-relevant but weaker than direct equity claims.

The stance mix is the finding. Across the 402 entity-level claims: 258 bullish, 78 neutral, 66 bearish, from 41 distinct sources, spanning 2011-06-23 to 2026-08-01. That is 16.4% outright bearish on a name most people would call consensus-long — and the bear claims are recent, specific, and from high-skill voices.

The bull lane — capital spending as investment:

> 2026-07-28 · anthony_pompliano · bullish · conviction 68

> "Google is intentionally betting the farm on AI capex (Q2 FCF turned negative), but it has a huge balance sheet; the ROI will come — maybe bigger than expected — and they'll look like geniuses."

> 2026-07-25 · anthony_pompliano · bullish · conviction 60

> "A forensic-accountant model put Google's capex bet at ~75% probability of working; cloud revenue up 80% YoY, backlog from $100B to $514B — they have no choice but to spend."

> 2026-07-25 · jordi_visser (skill 2.0) · bullish · conviction 74

> "Not worried about hyperscaler debt — low debt/equity, ~$2T of RPOs and ~$2T of compute already contracted ...; demand is already there and CapEx keeps accelerating."

> 2026-07-21 · compound_and_friends · bullish · conviction 72

> "Alphabet is the last clean AI hyperscaler; its earnings are the make-or-break moment for tech."

> 2026-03-27 · all_in · bullish · conviction 85

> "Google best positioned: existential motivation, agentic advantage via existing access to calendar/docs/email, and free cash flow to run separate consumer + GCP enterprise plays; most durable."

The bear lane — capital spending as malinvestment:

> 2026-07-23 · eurodollar_university · bearish · conviction 55

> "Google's exploding capex (44.9B in Q2, 2027 budget nearing ~400B per JP Morgan) is outrunning returns — each new dollar buys less compute as input prices rise; not a default call but returns are questionable."

> 2026-08-01 · eurodollar_university · bearish · conviction 55

> "AI-buildout debt is drawing growing caution — CDS on Oracle (215bps), SpaceX, Nvidia, Meta, Amazon, Alphabet moved higher in July."

> 2026-04-12 · jordi_visser (skill 2.0) · bearish · conviction 70

> "Magnificent Seven names already had their run and can't deliver the multiples investors need."

> 2026-06-06 · jordi_visser (skill 2.0) · bearish · conviction 68

> "Doesn't think the S&P 500 is safe past 2030 because AI will disrupt all companies; specifically Apple, Google and Amazon are not safe."

> 2026-04-06 · forward_guidance · neutral · conviction 55

> "Google is the best of the software bunch, but ad-based revenue loses value once agents, not humans, make purchase decisions."

And the most useful claim in the file, from a source that had been long and is no longer:

> 2026-04-14 · lyn_alden (skill 1.1) · neutral · conviction 65

> "Was long Google when cheap: AI takes non-monetized searches not ad searches, fortress balance sheet, YouTube benefits from Jevons video paradox — thesis now played out after doubling."

Three things this lane establishes.

First, the bear case is a returns-on-capital argument, not a business-quality argument. Not one bearish claim disputes that Alphabet's revenue is growing or that its cloud business is winning. Every one of them disputes whether the capital being spent will earn its cost. That is precisely the question Section 5 poses arithmetically, and it means the qualitative and quantitative lanes agree on what to watch.

Second, the single highest-skill voice in the file is on both sides within eight weeks — bearish on 2026-04-12 and 2026-06-06, bullish on 2026-07-18 and 2026-07-25. That is not incoherence; it is a voice distinguishing "the multiple has run" from "the debt is not the risk." But it means we cannot claim a strong directional signal from the deepest lane we track, and we do not.

Third, one tracked claim independently corroborates a filing figure to the dollar. The 2026-07-23 bearish claim cites capital expenditure of "44.9B in Q2"; the 10-Q filed the same day states capital expenditures "were $44.9 billion for the three months ended June 30, 2026." That is a useful validation of the ingestion pipeline and of the claim's specificity.

Management-lane claims, half-weighted, for completeness. From the 2026-04-29 investor-relations lane: a cloud backlog "nearly doubled QoQ to over $460 billion"; a 5% dividend increase to $0.22 quarterly (confirmed by the 10-Q); paid subscriptions reaching 350 million; and the autonomous unit "surpassed 500,000 fully autonomous rides per week." These are company statements and are treated as such — the dividend figure is independently verified in the filing; the backlog and subscription figures are not, and are used only as directional context.

Why Moderate rather than High. Breadth is exceptional — 41 sources, 402 claims, fifteen years. Direction is not. A lane that is 64% bullish, 19% neutral and 16% bearish, with its most skilled voice on both sides, supports a Hold better than it supports a Buy, and we let it.

8. Data integrity — what we rejected from the vendor file and why

Nine material items. Four are direct filing-versus-vendor contradictions, and in every one the filing wins.

1. FILING OVERRIDES VENDOR — liquidity. The vendor's newest balance sheet (FY2025, 2025-12-31) shows cash and short-term investments of $126.843B. The 10-Q filed 2026-07-23 states "As of June 30, 2026, we had $242.5 billion in cash, cash equivalents, and short-term marketable securities." The vendor understates liquidity by $115.7B. We use $242.5B.

2. FILING OVERRIDES VENDOR — debt. The vendor shows long-term debt of $46.547B and total debt of $59.291B at FY2025. The 10-Q states "As of June 30, 2026, we had long-term debt outstanding with a total carrying value of $98.2 billion", after $20.0B of US-dollar and $31.8B of foreign-currency senior note issuance in the first half. The vendor understates debt by roughly $51.7B. We use $98.2B.

3. FILING OVERRIDES VENDOR — the entire capital structure. The vendor payload contains no record of the June 2026 equity raise ($20.5B underwritten offering plus a $10.0B private placement), the $19.0B of 6.25% mandatory convertible preferred, the $40.0B at-the-market programme, or the complete suspension of buybacks. All are disclosed in the 10-Q. Any per-share or leverage metric in the vendor file predates a roughly $50B change in the capital base and a new preferred dividend obligation of ~$924M a year.

4. FILING OVERRIDES VENDOR — segment labels and a spurious row. The vendor labels the FY2025 $48.030B revenue line "Google Inc.", an opaque label; the same line in FY2024 reads "Google Subscriptions, Platforms, And Devices" and the 10-Q uses that category name. We use the filing's label. The vendor also carries an "Other Segments" row of −$127M in FY2025; the six real lines sum exactly to the reported $402.963B, so the extra row is a double-count and is discarded.

5. Trailing GAAP earnings and every GAAP-derived margin are economically meaningless — REJECTED as a lens. Q2'26 net income of $112.193B on operating income of $40.770B reflects $99.0B of unrealised equity-securities gains per the 10-Q. Consequently: priceToEarningsRatioTTM (18.75x) and priceToEarningsDilutedRatioTTM (18.97x) are rejected; ebitMarginTTM (67.4%), ebitdaMarginTTM (73.0%), pretaxProfitMarginTTM (67.1%), netProfitMarginTTM (54.8%), returnOnEquityTTM (50.8%), returnOnAssetsTTM (26.5%) and earningsYieldTTM (5.32%) are all rejected as contaminated. ebtPerEbitTTM of 2.026 is retained solely as evidence of the defect. This dive values on FY2027 consensus EPS of $14.994.

6. est netIncomeAvg is corrupt in the FY2026 row — REJECTED. The 2026-12-31 estimate row reports netIncomeAvg of $247.773B against ebitdaAvg of $183.714B — net income above EBITDA, which is impossible. The epsAvg figure of $20.259 is internally consistent with the quarterly consensus reconciliation and is retained as a contaminated number, clearly labelled; the EBITDA, EBIT and net-income estimate rows are rejected throughout and no EBITDA-based multiple appears anywhere in this dive.

7. netDebt ignores short-term investments — a $96.1B omission. Even on the vendor's own FY2025 data, netDebt of $28.583B is total debt ($59.291B) less cash and cash equivalents only ($30.708B), discarding $96.135B of short-term investments. Correctly computed, FY2025 was net CASH of $67.552B. km_ttm.enterpriseValueTTM of $4,627.207B carries the same error; rebuilt from the vendor's own inputs it is $4,487.659B, a 3.0% residual. All EV-derived multiples are rejected (evToSalesTTM 10.38x, evToEBITDATTM 14.21x, netDebtToEBITDATTM 0.175x, enterpriseValueMultipleTTM 14.21x).

8. Two internal vendor inconsistencies, flagged and resolved. (a) quote.marketCap is $4,555.211B while km_ttm.marketCap is $4,570.362B — a $15.15B discrepancy; we use the quote figure, which is consistent with the closing price. (b) quote.yearHigh of $404.47 and yearLow of $194.467 are intraday extremes; tech.hi52 of $399.04 and tech.lo52 of $195.32 are closing extremes. We use the closing series throughout and state the difference once. (c) inc_q reports Q4'25 revenue of $113.896B while earn_cal reports $113.828B for the same print — a $68M difference; immaterial, but noted rather than smoothed.

9. Headcount is stale. The profile reports 190,820 employees; the 10-Q states 198,933 as of 2026-06-30. The filing wins.

Not defects, correctly reported: freeCashFlowYieldTTM of 1.17% and priceToFreeCashFlowRatioTTM of 85.8x look absurd for a company of this quality but are correct — they reflect the genuine collapse in free cash flow documented in Section 5, and we use them with the explanation attached. dividendYieldTTM of 0.225% and lastDividend of $0.85 are consistent with the filing's $0.22 quarterly rate as it steps up from $0.21.

Non-equity tripwire: checked and clear. GOOG is common stock (Class C capital stock), beta 1.247, no fixed dividend, average volume 22.1M shares, and a 52-week range of $195.32-$399.04 — a 104% band. Nothing resembling a preferred or baby-bond profile. (Note that Alphabet has itself now issued a listed preferred instrument — the Series A and B depositary shares — but that is a different security from the one this dive covers.)

9. Technicals — a great year already banked

• Price $375.35, up 0.77% on the day, on volume of 19.7M against a 22.1M average — an unremarkable session.
• −5.9% from the 52-week closing high of $399.04, and +92.2% above the 52-week low of $195.32. Maximum drawdown from peak over the trailing year is only −5.9% — this stock has essentially not had a correction in twelve months.
• Above both moving averages: +5.3% over the 50-DMA of $356.29 and +14.9% over the 200-DMA of $326.76, both rising.
• RSI 52.1 · MACD −0.03. Both are as close to neutral as the indicators get. There is no momentum signal here in either direction.
• Relative performance: 12-month +97.6% against SPY +24.3% and QQQ +30.8% (a 67-to-73 point outperformance, the largest in this batch); 6-month +8.8% against SPY +11.1% and QQQ +15.7% (now lagging); 3-month −1.1% against SPY +7.6% and QQQ +7.7% (an 8.7-point deficit).
• The shape of that sequence is the whole technical story: a spectacular twelve months, of which the last three contributed nothing. The name led, then stopped leading, and has been consolidating since roughly May.
• Practical read: there is no technical urgency in either direction. The 50-DMA at $356.29 is the operative support and is 5.1% below; the 52-week high at $399.04 is 6.3% above. An investor who does not own this can wait without cost; an investor who owns it has no technical reason to sell. That is the definition of a Hold, and the chart agrees with the arithmetic.

10. Insiders — nothing to read, and we will not pretend otherwise

The honest reading is that this table contains no information. Eight transactions, two people, a total of 82 and 449 shares actually sold on the open market, and the remainder are stock-unit conversions and in-kind tax withholdings — the mechanical residue of equity compensation. No director or officer bought a share; none sold a meaningful position. For a company with a $4.6 trillion market capitalisation, insider activity of this scale is noise and is treated as such.

The transaction that actually matters is not in this table. The 10-Q discloses that on 2026-06-04 Alphabet completed a $10.0 billion private placement of 14 million Class A and 14 million Class C shares to an affiliate of Berkshire Hathaway Inc. That is not an insider transaction and does not appear in the insider feed, but it is by an order of magnitude the most significant ownership event of the period — a $10B commitment at roughly $353 a share, 6% below today's price. We note it as a fact from the filing and decline to over-interpret it: it is one buyer's decision at one price, disclosed by the issuer, not a signal we can weight.

The counter-signal is management's own. Alphabet sold $30.5B of common equity at $351-355 in June and repurchased nothing, with $69.5B of buyback authorisation available. A company that will issue stock at $353 and will not buy it at $375 is expressing a view.

11. Verdict, kill-criteria and flip conditions

Hold.

This is one of the finest businesses in existence, growing faster than it was three years ago, and it is not cheap enough to buy today.

What is genuinely excellent. Revenue accelerating from +8.7% (FY2023) to +24.2% (Q2'26) at a $400B+ scale. Google Cloud growing 82% year over year with operating income up $6.0B. A brand-new revenue line from selling the company's own accelerator systems, first recognised this quarter. Gross margin at a series high of 61.6%. Traffic-acquisition costs falling on favourable mix. $242.5B of liquidity and 296x interest coverage per the 10-Q. Return on invested capital of 15.1%. Forty-one distinct tracked sources discussing it, 258 of them bullish.

What holds it at Hold. The stock has risen 97.6% in twelve months against a market up 24.3%, and the last three months delivered −1.1% — the easy money has been made. The honest multiple is 25.0x FY2027 consensus, not the 18.9x every screen shows, because trailing earnings are inflated by $99.0B of unrealised marks on private equity stakes. Free cash flow has turned negative. The buyback has been switched off entirely while $30.5B of common equity and $19.0B of 6.25% preferred were sold — at prices below today's. There is $85.2B of not-yet-commenced data-centre lease obligation in no leverage ratio anywhere. And our own knowledge base, the deepest lane we track, is 16% outright bearish with its highest-skill voice on both sides.

Base fair value of $402 is 7.1% above spot, against a 28% bear case. That is not a buy. It is a fine thing to own.

Pre-registered KILL criteria — what would move this to Avoid:

• Google Search & other revenue growth falling below ~10% year over year for two consecutive quarters. It grew 13.4% in FY2025 and added $9.1B in Q2'26. This line is 55.7% of revenue and funds everything; its deceleration would simultaneously validate the agent-disintermediation bear case and remove the cash flow behind the capital programme.
• Google Cloud operating margin declining as the 2026 depreciation flows through — the direct test of whether $80.6B of half-year capex was investment or expense.
• Drawdown on the $40.0B at-the-market equity programme, which would signal that $49.5B of June issuance plus $51.8B of notes was insufficient.
• Free cash flow remaining negative through FY2027 while capital expenditure continues to rise.
• An adverse antitrust remedy affecting the distribution agreements underpinning search economics, beyond the $5.2B European fine already paid in July 2026 and the $2.1B Swedish accrual taken in Q2'26.
• A reversal of the equity-securities marks. The $99.0B gain that made Q2'26 look extraordinary can run the other way; it is not cash and it is not the business, and a large negative mark would be equally meaningless — but it would hit reported book value and the screens would react.

Pre-registered FLIP TO BUY (upsize) criteria:

• A price below ~$330 (22x FY2027E), which would restore a genuine margin of safety and roughly coincide with the 200-DMA of $326.76.
• Resumption of share repurchases against the $69.5B authorisation — the clearest available statement that management considers the capital programme funded and the shares undervalued.
• Google Cloud operating margin expanding for two consecutive quarters through the depreciation ramp.
• Accelerator-system revenue disclosed as a separate, material line with third-party customers named, converting an internal cost advantage into an external franchise.
• Free cash flow returning to positive while revenue growth stays above 18%.

Where GOOG fits in the Synthos Framework Portfolio. The mega-cap platform sleeve, as a held position rather than a new one — if already owned, a 2-3% weight is defensible on business quality alone; if not owned, there is no reason to initiate here. This is a name to accumulate on weakness toward the 200-DMA, not at 6% below a 52-week high after a near-double. Note the house standing rule that mega-cap incumbents are guilty until proven innocent on a forward basis: Alphabet largely proves its innocence on growth (+24.2% revenue, cloud +82%) and fails it on price (25.0x FY2027E after +97.6%).

On share classes. GOOG is the Class C capital stock, which carries no voting rights; Class A (GOOGL) carries one vote and Class B, held by founders, carries ten. The two listed classes are the same company and the same economics, and this dive is struck against GOOG's own close of $375.35. Where a Class A dive exists it is written independently; nothing here assumes or depends on it.

Single biggest risk: the return on the capital programme. $80.6B spent in six months against $39.6B in the comparable half, financed by $51.8B of debt and $49.5B of equity and preferred, with $85.2B of lease commitments not yet begun and free cash flow already negative. The depreciation from this arrives in the income statement before the revenue does, and consensus already models zero net margin expansion for four years as a result. If the demand behind it is as contracted as the bulls argue, this is the best capital allocation of the decade. If it is not, Alphabet has converted a fortress into a leveraged infrastructure business at exactly the wrong multiple. Our knowledge base contains both arguments, made recently, by credible people, and we do not claim to know which is right.

Most fragile assumption in the price: that advertising economics are structurally durable through a shift to agent-mediated purchasing. Google Search & other is 55.7% of revenue and pays for everything else. The most specific bear claim we track argues that "ad-based revenue loses value once agents, not humans, make purchase decisions." Nothing in the current numbers tests that claim — Search grew $9.1B in the June quarter — which is precisely what makes it fragile: it is unfalsifiable on a quarterly cadence, and so it is not priced.

Provenance & disclosures

• Traceability: 402 entity-level knowledge-base claims name Alphabet (510 raw hits; 108 thesis-only mentions excluded; breadth 41 distinct sources; net conviction mixed). Stance mix 258 bullish / 78 neutral / 66 bearish, spanning 2011-06-23 to 2026-08-01. Ten claims tagged to a quarantined-misattribution source were excluded outright under the standing attribution policy; management-lane claims (chief executive, research unit and investor-relations lanes, skill 0.5) are half-weighted and identified as such wherever cited. Searches covered GOOG/GOOGL (case-sensitive), Alphabet, Google, the chief executive's name, and four product-level terms covering the research unit, the assistant product, the autonomous-transport unit and the video platform, across all 51,928 records. All quoted claims are verbatim from the stored claim text.
• Data as-of: fundamentals — income statement 2026-06-30 (Q2'26, filed 2026-07-23, twelve days before this dive); annual balance sheet and cash flow 2025-12-31 (FY2025, filed 2026-02-05), with all material balance-sheet figures superseded by the 2026-07-23 10-Q as documented in Section 8 · estimates 2026-08-04 · prices 2026-08-04 ($375.35; 50-DMA $356.29; 200-DMA $326.76; RSI 52.1) · knowledge-base claims 2026-08-04. Every figure comes from the Synthos vendor data file for GOOG or from the SEC filings in filings/GOOG/; no figure comes from memory, recall or external retrieval.
• Filings consulted: 10-K filed 2026-02-05; 10-Q filed 2026-04-30; 10-Q filed 2026-07-23 (the primary source for this dive); 8-K filed 2026-06-05; 8-K filed 2026-06-11 (the preferred-stock listing); 8-K filed 2026-07-22.
• Filing-versus-vendor overrides (four, all resolved in favour of the filing): (1) liquidity — vendor $126.843B, 10-Q $242.5B, we use the 10-Q; (2) long-term debt — vendor $46.547B, 10-Q $98.2B, we use the 10-Q; (3) capital structure — the vendor payload records none of the $20.5B underwritten equity offering, the $10.0B private placement, the $19.0B of 6.25% mandatory convertible preferred, the $40.0B at-the-market authorisation, or the complete suspension of buybacks, all disclosed in the 10-Q; (4) segment labelling — the vendor's "Google Inc." row is the filing's "Google subscriptions, platforms, and devices", and the vendor's −$127M "Other Segments" row is a spurious double-count. Headcount is a fifth, minor override: vendor 190,820 against the filing's 198,933.
• Other data rejections (detailed in Section 8): trailing GAAP price-to-earnings (18.75x/18.97x) rejected as a lens because Q2'26 net income includes $99.0B of unrealised equity-securities gains per the 10-Q; ebitMarginTTM, ebitdaMarginTTM, pretaxProfitMarginTTM, netProfitMarginTTM, returnOnEquityTTM, returnOnAssetsTTM and earningsYieldTTM all rejected as contaminated by the same gains; est FY2026 netIncomeAvg ($247.773B) rejected as exceeding ebitdaAvg ($183.714B), and all EBITDA/EBIT estimate rows rejected throughout; netDebt corrected from $28.583B to net cash of $67.552B on the vendor's own FY2025 data (omits $96.135B of short-term investments), and enterpriseValueTTM corrected from $4,627.207B to $4,487.659B, with all EV-derived multiples rejected; quote.marketCap ($4,555.211B) preferred over km_ttm.marketCap ($4,570.362B); closing-price 52-week extremes preferred over intraday; a $68M mismatch between inc_q and earn_cal Q4'25 revenue noted.
• Estimate coverage: FY2026 32 analysts on EPS and 29 on revenue; FY2027 34 on EPS and 31 on revenue — the primary forward anchor; FY2028 27 and 14; FY2029 30 and 26; FY2030 16 and 15. No conclusion in this dive rests on the FY2028 or FY2030 revenue rows (14 and 15 analysts respectively). The FY2026 EPS row is used only to demonstrate contamination, never as a valuation input.
• Peer note: the vendor peer set contains seven genuine mega-cap technology comparables (Apple, NVIDIA, Microsoft, Broadcom, Taiwan Semiconductor, Meta, Cisco) and three that do not belong (ExlService $5.2B, Flex $46.4B, Garmin $59.3B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $270 / $402 / $510 anchors are multiples of FY2027 consensus EPS of $14.994 (18.0x / 26.8x / 34.0x), cross-checked against FY2028E of $17.679. Stated arithmetic, not a discounted cash flow. The base case assumes the multiple broadly holds and earnings compound into it; it does not require re-rating. The bull case does require re-rating and is flagged as the fragile leg.
• Timing: Q2'26 was reported 2026-07-22 and the 10-Q filed 2026-07-23, twelve days before this dive, so the most recent quarter is fully reflected. The next print is 2026-10-28, 85 days away — no earnings banner applies. That print is the first clean quarter (consensus EPS $3.02) and will be the first opportunity for the market to see the underlying earnings power described in Section 2.
• Accessibility note: all tables are labelled, all emphasis is carried by text rather than colour, and every figure is stated in full rather than encoded in a chart.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.