Alphabet — The Convention Price

Two ways to count the same company

Report date 2026-07-24 · Data as-of 2026-07-23

An advertising business buying an infrastructure company

Most people think of Alphabet as Google, and Google as search. That was a fair description five years ago. It is no longer one. In the three months to June 2026, Alphabet's advertising businesses — Search, YouTube and its ad network — produced $81.6bn of the company's $119.8bn of revenue and almost all of its profit. Google Cloud produced $24.8bn, growing 82% year on year. The company spent $44.9bn building data centres in that single quarter, which was more than the entire $39.1bn of cash its operations generated over the same three months.

One quarter's data-centre bill exceeded one quarter's operating cash. Alphabet is an extraordinarily profitable advertising business pouring every dollar it earns, plus a great deal that it has borrowed, into becoming an infrastructure company. For the full year it plans to spend between $195bn and $205bn on capital expenditure — buildings, chips, power. Three years ago that number was $32.3bn. Management has said 2027 will be "significantly higher" and has not given a figure. Everything in this report is, in one way or another, a question about that spending: does it earn its keep?

The operating business performed. The capital plan changed.

Data shows the operating business performed. Q2 2026 revenue grew 24.0% to $119,796M. Operating income was $40,770M at a 34.0% margin, against 32.4% a year earlier. Google Cloud grew 81.8% to $24,768M and its operating margin went from 20.7% to 35.6% in a year. The Cloud order book stands at $514bn, up more than $50bn in a single quarter. Search grew 16.8% on record query volume. This is not a broken company.

Data also shows a capital pivot, in the same filing. In the June quarter Alphabet spent $44,924M of capital expenditure against $39,069M of operating cash flow, so free cash flow for the quarter was negative $5,855M. It repurchased $0 of its own stock, against $13,238M in the prior-year quarter and $45,709M across FY2025. In the first half it borrowed $56,226M, sold $49.6bn of new common stock and mandatory convertible preferred, and registered a further $40bn at-the-market equity facility, undrawn at 30 June 2026 and primarily intended to meet tax obligations on employee equity grants. Long-term debt roughly doubled, from $46,547M at 31 December 2025 to $98,165M, and the share count rose rather than fell.

None of this is hidden and none of it is improper. But an investor who owned Alphabet because it was a machine for buying back its own stock now owns something different.

And the headline number flattered all of it. Reported diluted EPS for the quarter was $9.11, up 294% from $2.31. Of that, $6.26 — 69% — came from the $99,031M pre-tax gain on equity securities: $77,354M of marks on privately held stakes that have no market price, plus $21,399M on marketable and other holdings, against $278M actually realised. Alphabet attributes the gain primarily to unrealised gains from SpaceX and one other private company. It is a real gain in the accounting sense. It is not money the business earned, it cannot be spent, and it can reverse.

The market pays the convention price

Three conventions, and the gap closes

There is a technique called a reverse discounted cash flow — instead of forecasting a company's future and calculating what it is worth, you take today's price and work backwards to find what the market must be assuming. We expected to find heroic growth. We did not. Solved one variable at a time and holding our own conventions, the pre-print close of $342.09 requires a 22.2% six-year revenue CAGR ($1.66tn of 2032 revenue), or a 58.3% terminal EBITDA margin, or an 8.14% cost of equity (a 2.79% equity risk premium), or 6.0% perpetual growth. None is individually defensible, which usually means the model is being asked the wrong question.

So we asked a different one. Take our base case — free cash flow to equity, the cash left for common shareholders after operating costs, tax, capital expenditure, working capital and net debt movements — and apply, one at a time, three conventions professional analysts customarily use and we do not. The walk is short, and it ends where the market already is.

We believe this is the whole argument. An investor's answer to two questions — "is stock compensation a real expense?" and "what return should I require on a $3.9tn equity?" — determines everything else. We think employee share grants are a genuine cost and that a company of this size should be discounted at the rate the standard model produces, not the rate that makes the answer comfortable. On that basis Alphabet is worth less than it trades for, and we say so. But an investor should understand precisely what they are disagreeing with us about. It is not Alphabet's future. It is arithmetic convention.

What is actually free about the free cash flow

The same convention shows up in the cash-flow statement, and it is the second number we would put in front of a reader. Alphabet's reported trailing free cash flow is $53,273M. That figure is struck after adding back $28,147M of stock-based compensation — 53% of the total. Deduct it, as this report does throughout, and genuinely free cash is $25,126M, which the $3,885bn market capitalisation at the anchor close covers 155 times, and the $4,184bn capitalisation at the pre-print close 167 times.

Meeting the $195,000–205,000M full-year capital-expenditure guide — raised from $180–190bn, roughly 40.3% of revenue — requires roughly $119bn of spending in the second half alone.

The depreciation wave arrives in every scenario

A fourth force runs underneath the whole model. When a company spends $200bn on equipment, the accountants spread that cost across the following years as depreciation. Alphabet's depreciation and amortisation charge is currently about 5.8% of revenue. In every version of the future we model — good, middling and bad — it rises to somewhere between 11.9% (base) and 15.8% (bear) by 2032. That means operating margin, measured properly, falls in every scenario, including the ones where EBITDA margin expands: base-case EBIT margin goes from 34.7% in 2026E to 30.8% in 2030E. First-half 2026 D&A already grew 43% year on year, and net property and equipment stands at $321,212M, 72.0% of trailing revenue, against 48.9% in FY2024. Return on invested capital has already fallen from 41.3% in FY2024 to 36.2% in FY2025 to 33.9% trailing — though the spread over a 10.14% WACC remains a wide +23.7 percentage points.

Investors anchored on cash-flow-style measures will not see this coming. Investors anchored on earnings will.

Where this note sits in the series — "The Benchmark"

Added 30 July 2026 (v1.1). The valuation, scenarios and fair value below are unchanged from the 24 July original; this section connects the note to the house depreciation framework it now travels with.

This note is a satellite of The Depreciation Question (17 July) and its Q2 scoreboard, The First Invoice (30 July). In that framework Alphabet is the benchmark: the gentlest case of the group's stretched-clock problem, and therefore the yardstick the others are measured against. It runs the lowest capex-to-income ratio of the four hyperscalers (0.69× in FY2025, against Meta's 1.15× and Amazon's 1.70×), carries the mildest earnings re-mark on the house three-year server clock — FY2025 EPS $10.81 reported against $10.35 restated, a 4.4% overstatement, versus 9–17% elsewhere — and it is the only company of the six that discloses a clean point-estimate server life ("generally six years") rather than a range.

The Q2 filings sharpened both halves of that description. The benchmark held its discipline: the 10-Q discloses no change in estimate, so the six-year clock stands, and the 60%-servers footnote our share estimate rests on is restated unchanged. And the benchmark is drifting: first-half depreciation rose 43%, first-half capex more than doubled to $80.6bn, the capex-to-income ratio moved from 0.69× toward roughly 1.2× (income measured excluding the securities gains discussed above), and $122.8bn of property and equipment now sits "not yet in service" — bought, owned, and not yet depreciating — up from $78.6bn six months earlier. That queue is tomorrow's depreciation wave, and it is why the wave section above rises in every scenario we model.

One point of accounting honesty cuts both ways. Our fair value here is built on cash flows, and a depreciation schedule does not change cash — so the house re-mark does not move the $256.42. What it moves is the quality of the reported earnings the market is capitalising: the convention price of the title is being paid on a bottom line that the series' own arithmetic marks as roughly four to five per cent flattered, before the wave crests. Within the series, Alphabet is the name where that gap is smallest — which is precisely what makes it the benchmark against which Meta, Amazon and Microsoft are read. (The role of pure control belongs to Apple, which sits outside the buildout entirely; Alphabet is the yardstick inside it.)

The counter-evidence, stated plainly

We are not arguing that the business is broken. The evidence against our conclusion is set out at full strength later, under the case against us, rather than summarised into weakness here. In short: the capital programme has a $514bn order book behind it; Cloud's incremental operating margin was 53.7% while cost of revenue grew 17.7% against 24% revenue growth; and Cloud's growth is substantially organic — Wiz, consolidated from 2026-03-11, contributed about 2% of Cloud's $11,144M year-over-year increment, leaving ex-Wiz growth near 80% against the reported 81.8%.

The offsetting fact is the interface rather than the machine room. AI Overviews appear on roughly 18% of queries and 57% of long-tail queries; within those results, organic click-through on informational queries fell from 1.76% to 0.61% between June 2024 and September 2025 — a 61% decline — per Seer Interactive, published November 2025, measuring 3,119 informational queries across 42 organisations. That is a change over time inside AI-answer results, not a comparison of AI answers against ordinary ones, and it does not cover commercial queries. On assistant share, Momentic put ChatGPT at 54.7% of worldwide assistant web visits in June 2026 against Gemini's 27.4%; Sensor Tower, counting users rather than visits, put the split at 46.4% to 27.7% in May 2026. We believe Alphabet is winning the machine room and losing, for now, the front door — and that so far the front door has cost it nothing measurable.

Three methods, one balance

Three methods, weighted by how much independent information each carries. The intrinsic method takes half the weight because the pivotal question — whether $1.3tn of cumulative capital expenditure earns its cost of capital — is a cash-flow question that relative multiples silently assume the answer to. The two relative lenses are capped at 50% combined because both inherit the market's own conventions and so cannot independently test the thesis.

The payoff is skewed the wrong way

We believe the shape of this payoff matters as much as its centre. The case against runs to a −51.6% de-rating, four times the reward on the other side, and no probability weighting makes that attractive. For completeness on where we sit against others: the Street mean target is $433.55, with 57 of 64 covering analysts at Buy or Strong Buy, 7 at Hold and none at Sell (analyst actions dated through 2026-07-23). It sits above every method here; we report it neutrally and it does not move our number.

What would change our mind

This report is one variable away from a different conclusion, and the variable is named. Terminal capital expenditure as a share of revenue is worth $77 per share across its plausible range — $173 at 22%, $250 at 13% — and it is the input on which management has given the least guidance beyond 2027. It carries the highest ratio of value-at-stake to available information anywhere in the model. The FY2027 capital-expenditure guide, expected with FY2026 results in late January or early February 2027, is the single disclosure that resolves it. For scale, the other two levers are revenue CAGR at $85 per share across its range and terminal EBITDA margin at $65.

Two showings would close most of the gap between our fair value and the price: FY2027 capex implying capital intensity below 35% of revenue with a stated glide path, or Cloud backlog converting at or above the 50%-in-24-months rate with segment margin holding above 30% after capacity eases. The arithmetic turns constructive at $192 or below absent a corresponding deterioration in the operating KPIs — that is, on a de-rating in convention rather than in fundamentals; $192 is a 25% margin of safety to fair value and coincides with the $194 Monte Carlo median of the intrinsic distribution. Conviction hardens the other way if Search & other growth prints below 10% for two consecutive quarters, if the DC Circuit broadens the search-data sharing remedy, or if FY2027 capex is guided above roughly $260bn. And we concede the thesis outright if revenue CAGR tracks above 16% while terminal EBITDA margin holds above 44%: in that world the FCFE value is $334 and the argument was about conventions that had ceased to matter.

The case against this ledger

Likelihood and magnitude are each scored 1–5 over a twelve-month horizon; score is the product. Likelihood: 1 = <10%, 5 = >75%. Magnitude: 1 = <$10 per share of fair value, 5 = >$100. The five highest-scoring risks are shown; the full thirteen-risk register is in the annex.

Next in the register, and material: an adverse DC Circuit ruling broadening the December 2025 search-data sharing remedy (2 × 5 = 10); Cloud margin mean-reverting once capacity ceases to be the binding constraint (3 × 3 = 9); and dilution from the $40bn ATM plus the loss of a roughly $45bn annual buyback bid (4 × 2 = 8).

The steelmanned bull case

The capital programme has an order book behind it. A $514bn backlog, up more than $50bn sequentially, with capacity rather than demand as the stated constraint. If it converts at the pace management implies, the spending is underwritten by contracts rather than by hope.

The unit economics are improving while it converts. Cloud's incremental operating margin was 53.7% and cost of revenue grew 17.7% against 24% revenue growth. If that persists, the depreciation wave lands on a business whose gross economics are getting better, not worse — and our terminal margin assumption of 44% is too low.

And the franchise has not broken, while the silicon is genuinely rare. Search grew 16.8% on record queries with traffic acquisition cost falling to 19.8% of advertising revenue from 20.6%, and Alphabet owns the only fully integrated silicon-to-model-to-distribution stack among US-listed companies, with the majority of contracted third-party TPU revenue expected in 2027. We concede the thesis if revenue CAGR tracks above 16% with terminal EBITDA margin above 44%: the FCFE value is then $334 and the convention argument stops mattering.

Disclosures & Data Integrity

Report date 2026-07-24 · Data as-of 2026-07-23

Price and date convention

Report date: 24 July 2026. Data as-of 23 July 2026 (snapshot 2026-07-23-134714). Two price references are used throughout and are named consistently. The anchor close is $317.69, the 23 July 2026 close and the last completed session at the time of analysis; the recommendation and every fair-value comparison are struck against it. The pre-print close is $342.09, the 22 July 2026 close and the last session before the Q2 2026 release; it is the reference for the convention bridge and for the trailing multiples quoted at the pre-print price. 52-week range $187.82–$408.61, intraday, over the 52 weeks ended 23 July 2026.

Sources, assumptions and limitations

Every figure in this report traces to a dated source in the manifest of the 23 July 2026 data capture. Primary sources are Alphabet's Q2 2026 Form 8-K Exhibit 99.1 (accession 0001652044-26-000066, filed 2026-07-22) and Form 10-Q (accession 0001652044-26-000071, filed 2026-07-23), the FY2025 Form 10-K (accession 0001652044-26-000018, filed 2026-02-05) and prior 10-Ks, the DEF 14A (accession 0001308179-26-000342, filed 2026-04-24), Form 4 insider filings, verified end-of-day price data and a dated media sweep for the independently reported rows. Where a figure is derived rather than disclosed it is labelled as such and its method stated; where a needed disclosure does not exist it is flagged as a gap rather than estimated. The full source-and-evidence log, the key-assumptions log, the thirteen-row risk register, the disclosed limitations and the glossary are reproduced unabridged in the annex and are not summarised here.

Three limitations bear directly on figures a reader of this condensed edition will quote. First, the five-year clean-EPS percentile band is internally constructed — built from daily closes divided by the last-reported fiscal-year diluted EPS across 1,255 sessions, with the current placement using this report's own clean-EPS adjustment. No external data provider publishes a series that reproduces it; the 77th-percentile figure is a TON618 construction, not a vendor statistic. Second, the 4.5% equity risk premium inside the 10.28% cost of equity is an analyst assumption rather than a market-implied measurement, and it is among the inputs with the most leverage on the output; the annex runs the cost of equity across 9.50%–11.00%, a $53.34 per-share range, so the disagreement is arithmetic rather than rhetorical. Third, Alphabet does not disclose acquired revenue separately, so the inorganic contribution to reported growth is bounded rather than measured: Wiz had passed roughly $1bn of annual recurring revenue at close, about 2% of Cloud's $11,144M year-over-year increment, which puts ex-Wiz Cloud growth at approximately 80% against the reported 81.8%. Alphabet also does not disclose Cloud customer concentration, so the composition of the $514bn backlog is carried as an unquantified exposure.

This condensed edition selects from and does not add to the audited section set. No figure, method, scenario, probability or conclusion differs from the full analysis; where the full analysis carries a caveat on a figure, the caveat is carried here with it.

Standard disclosures

Information only. TON618 Capital. This report is for information purposes only. Nothing here is an offer to sell or a solicitation of an offer to buy any security, fund interest, or digital asset, and nothing here is personalized investment advice or a recommendation regarding any instrument.

Publisher's exclusion. All research is published solely as general, impersonal information of regular circulation. It is not tailored to the objectives or circumstances of any individual and is not issued in connection with compensation from any client. The Fund has no clients and distributes all research free of charge. On that basis it publishes in reliance on the publisher's exclusion from the definition of "investment adviser" under the Investment Advisers Act of 1940 (§202(a)(11)(D); cf. Lowe v. SEC, 472 U.S. 181 (1985)).

Registration & conflicts. TON618 Capital is not registered as an investment adviser or broker-dealer in any capacity. The Fund is a Bitcoin fund and may hold or transact in the securities or digital assets it discusses; material conflicts are disclosed where they exist. Ownership: the Fund holds no position in GOOGL as of the report date, and GOOGL is not a Fund holding or a BTC-correlated instrument. Compensation: the Fund received no compensation from any party in connection with this report and charges nothing for it.

Use of AI. Artificial intelligence is used in the creation of this research. All methodology and data integrity are reviewed and approved before publication by TON618 Capital's Chief Investment Officer, Keyth Beck; errors may nonetheless occur, and readers should verify independently.

CFA. This report was prepared to align with CFA Institute analytical standards (methodology only). CFA® and Chartered Financial Analyst® are registered trademarks owned by CFA Institute. That reference describes the analytical framework applied; it does not imply the report was prepared, reviewed, or authored by a CFA charterholder, and the report is not issued, reviewed, endorsed, certified, or approved by — nor affiliated with — CFA Institute.

Risk & feedback. Past performance is not indicative of future results. Digital assets and equities are volatile and may result in total loss of capital. Corrections and feedback are welcome — please direct them to CIO Keyth Beck at keyth@ton618capital.com. Version 1.0 · analyst: TON618 Equity Research.

Version and change log

Annex. The complete 144-page analysis — eighteen sections covering the business model, moat, proprietary technology, five-year financial statement analysis, unit economics, the gearing map, the operating forecast, the intrinsic and relative valuations, the full risk register, catalysts, ESG and governance, and the unabridged source, assumption and limitation logs — is available as the annex, on request.