How to Read the Alphabet Investment Case: A Plain-English Guide

A plain-English guide to Blockhaven's Alphabet (GOOGL) research: what the figures show, what Blockhaven calculates and assumes, and where judgment enters.

How to use this guide

The research contains the evidence, calculations, assumptions and full argument. This guide explains how to read it. As you go, ask whether each statement is a reported figure, a calculated one, an assumption or Blockhaven's judgment. A precise-looking number should not carry more authority than it has earned.

The Alphabet investment case in plain English

Alphabet is earning more than it ever has. Search is growing faster than a year ago, Google Cloud's profit margin has risen sharply and a contracted backlog says customers want a great deal more of it. To serve that demand the company is spending on equipment, data centers and chips on a scale it has never matched, and it has raised new equity and borrowed to do it.

The question the research asks is what the price already assumes. At $340.92 a share, the price is consistent with the new capital earning well above its cost. Our base case, a reference path in which Cloud demand fills the capacity the guided build creates and the modeled return on the added capital is only a little above the discount rate, is worth $178.92 from its cash flows, and the bull case, in which that modeled return is well above the discount rate, is worth $419.72. The base value is the conservative reference within a range we chose, not a measured or most-likely outcome: with fuller use of the capacity it is worth up to $208, the path on which demand fades after the build is worth $134, and a strong coherent stack gives up to $291, still below the price.

How Alphabet makes money

Alphabet earns its money in three ways. Google Services sells advertising on Search, on YouTube and, through the Network, on other companies' sites and apps, and also sells subscriptions, apps and devices. Google Cloud rents computing power and software to businesses and, since the second quarter of 2026, has also sold its own TPU chip systems to customers. Other Bets, mostly the Waymo robotaxi business, is a small revenue line that loses money.

In FY2025 Services produced $342.7 billion of revenue, 85.1% of the company, at an operating margin of 40.7%, and Cloud earned $13.9 billion on $58.7 billion of revenue. The mix is moving: in the second quarter of 2026 Cloud revenue was $24.8 billion, up 82% on the year, at an operating margin of 35.6%.

Two cautions help when reading these numbers. Segment margins are reported before the shared costs that sit outside every segment, mostly shared AI research, so each segment's margin would be lower if those costs were allocated to it; the consolidated margin of 32.0% in FY2025 is the better measure of what the whole business earned. And Services is still the profit engine, while Cloud is where the AI spending shows up as revenue most directly, which is why the research spends so much of its time on Cloud.

Why Search monetization matters

Search is the largest line in Alphabet's revenue, the heart of Google Services, the segment that earns most of the company's profit, and the place where the central question about AI is being answered in public. Search and other revenue grew 17% in the second quarter of 2026 to $63.3 billion. Monetization means turning a person's question into revenue, mostly through the ads shown beside the answer.

If AI answers keep people searching and advertisers paying, the business Alphabet already has stays strong, and the profit it produces is what pays for the AI spending. If AI answers instead reduce the clicks that ads depend on, or cost more to produce than they earn, the profit that funds the whole program is weaker. That is why a Search number matters to a reader who cares about Cloud or chips.

One cost deserves care. Alphabet pays partners for the places where people start a search, and it reports that as traffic acquisition cost: $16.2 billion in the quarter, 19.8% of advertising revenue against 20.6% a year earlier. The line also includes payments to Network partners, so a lower rate is not by itself proof that Search became cheaper to acquire. So far the reported revenue is passing the test, and the comparison gets harder from the third quarter, because a year ago was already the start of an acceleration.

How to read the Blockhaven Rating

Three different dimensions are easy to run together.

The five levels are 5/5 Compelling, 4/5 Favorable, 3/5 Balanced, 2/5 Cautious and 1/5 Poor. This is Blockhaven's own scale, not stars and not Buy, Hold or Sell. Alphabet is rated 2/5 Cautious, with low confidence and a thesis of "Constructive only under specific assumptions." The rating was assessed at $340.92 on September 29, 2026, over the horizon to the end of 2028.

The business is excellent and the rating is cautious, and there is no contradiction: the rating weighs what the price offers for the risk, and a very good business can be fully priced. A lower price could improve the balance without any change in the business.

Four ideas are easy to confuse. The base case is a reference path: a coherent set of assumptions drawn from reported run-rates, used to value the company. It is not a forecast of the most likely outcome and it carries no probability. Blockhaven's central judgment is what the evidence lets it say about the thing that decides the value, which is the profit the new capital will earn: the demand is visible, that profit has not been shown, and the price already pays for most of what the strongest case would deliver. Confidence is a separate question about how strong the evidence is, and it is not an expected return.

Reported, calculated, assumed and judgment

Every statement in the research belongs to one of four categories, and each keeps its own status.

A calculation can be exact and still rest on a debatable assumption. Precision and certainty are not the same thing.

Google Cloud: a backlog is not revenue

A backlog is contracted revenue that has not yet been recognized. It says customers have committed; it does not say when the revenue arrives or what it costs to serve.

At the end of the second quarter Alphabet's backlog was $519.5 billion in total, of which $513.9 billion was Google Cloud, which is 6.6x the last four quarters of Cloud revenue. The company expects to recognize just over half of the total within two years. Spread evenly over eight quarters, that is an average of about $32.5 billion a quarter for the total and at least about $31.8 billion for Cloud, against the $24.8 billion Cloud earned in the second quarter.

That is an average of an expectation and not a schedule. Recognition may start slowly and finish quickly, and contracts signed later add to Cloud revenue, so a quarter below the average does not mean the expectation has failed. It is strong evidence of demand and incomplete evidence of profit. Three questions decide how far to trust it: how fast it converts, at what margin, and how concentrated the buyers are. Alphabet does not say how concentrated its buyers are. A backlog owed mostly by a few companies that are themselves raising money to buy the capacity would be riskier than one spread across many customers, and that is a scenario to keep in view and not a fact.

Google Cloud: depreciation is not replacement, and useful life and utilization

When Alphabet buys equipment, the cost does not hit profit at once. It is spread over the equipment's assumed life as depreciation, here about six years for servers. Capital spending was $44.9 billion in the second quarter and depreciation $7.1 billion, so the quarter's profit carries only a small part of the spending that will depreciate for years.

That gap closes as the equipment goes into service. In the base case depreciation rises from $30.5 billion in 2026 to $252.9 billion in 2035, and capital spending in 2035 is $276.2 billion, which is why the margin falls in the base case even though revenue keeps growing. Depreciation is an accounting allocation of past spending and not a measure of what it costs to keep the business the same size. What replacement costs depends on how long the equipment really lasts and how much more each new dollar of it can do, and neither is reported. The base case assumes that the capital needed for a dollar of revenue falls 3% a year after 2029, and with that assumption spending in 2035 comes out at 1.09x depreciation. That is a modeled assumption and not a requirement.

Two facts decide whether a large equipment bill is a good investment: how long the equipment stays useful, and how fully it is used. A server that lasts six years and runs close to full earns its cost back over more revenue than the same server standing idle for part of its life. Utilization is the share of capacity that is actually earning. The base case assumes that demand grows until the capacity Alphabet is building is used at least as fully in 2029 as the earlier path used it in 2027, which puts the base case at 92% in 2029. On the earlier path, in which demand fades after the capacity is built, owned capacity is used only 79% as fully in 2029 and the same spending is worth $134 a share against $178.92 in the base case. Nothing in the record measures utilization after 2027, so this is an assumption and the value moves a great deal with it. Neither a longer life nor a fuller one is a fact until the company reports it.

Profit is not cash

In the second quarter, four numbers answered four different questions: net income of $112.2 billion, operating income of $40.8 billion, operating cash flow of $39.1 billion and free cash flow of -$5.9 billion. They can disagree by a wide margin.

Reported net income included $99.0 billion of gains on equity securities, before tax, that have not been sold. The company says the gains added $77.1 billion to net income after $21.9 billion of tax, and $6.26 to earnings per share. They are real accounting profit and not what the business earned; operating income of $40.8 billion is. Earnings per share were $9.11 as reported and $2.85 without the gains, which is the figure we use. Cash is a third answer: operating cash flow was $39.1 billion, capital spending $44.9 billion, and what remained, free cash flow, was negative. A company can report rising profit and still be a net user of cash while it builds.

Temporary tax effects

The 2025 tax law lets a company deduct the full cost of qualifying equipment acquired after January 19, 2025 in the year the equipment is placed in service and not the year it is ordered or paid for. While the equipment placed in service keeps growing, cash taxes are lower than the taxes in the income statement. The difference narrows gradually as the deductions for earlier equipment run off against the depreciation that follows. In the base case the timing difference is worth $6.28 a share to the firm, and our terminal value carries none of it.

The same law lowers the deduction for income from foreign sales from 37.5% to 33.34% from 2026. One effect helps for a few years and the other is permanent, which is why the research treats them separately and tests both in the sensitivity table.

Scenarios are not forecasts

A scenario asks what the value would be if its stated assumptions were true. It does not assign a probability. The research builds each scenario in four steps: business assumptions, an operating path to 2035, cash flows and earnings, and a value per share.

Bear, base and bull are if-then descriptions of the company, not assigned probabilities, and they are not the same as the stress below the bear case, which tests what a permanent impairment would leave. The first column is a value today. The other two are values on December 31, 2028, and only they can be compared with each other. The spread shows how much the result depends on one question: what the new capital earns after the depreciation it creates.

What a DCF value is, and what it is not

A discounted cash flow, or DCF, value is the worth today of the cash a business can pay out over its life, after the money it must spend to keep growing, translated into today's dollars with a discount rate. It answers one question: what is this stream of cash worth if these assumptions are true?

It does not say where the share price will be next month or next year. A price target is a forecast of a price on a date, and prices move for many reasons that cash flows do not capture. The base value of $178.92 is therefore not a price target, not a forecast of the most likely outcome and not a prediction that the shares will fall to it. What it says is that the price of $340.92 asks for more than the base assumptions deliver, and the bull value of $419.72 shows how much would have to go right.

Discount rate and hurdle rate

A dollar received in future is worth less than a dollar today because an investor must wait and take risk. Discounting translates future amounts into today's terms. The discount rate in the cash-flow method is 9.5%, built from the Treasury yield, an equity risk premium, a beta and the cost of debt. The illustrative hurdles of 10% and 12% used to judge the values at the end of 2028 are different: they are yardsticks, not statements of what any investor requires.

Two methods, two dates

The discounted cash flow runs to 2035 and adds a terminal value, and 76% of the base case's enterprise value sits in that terminal value because the early years consume cash. The earnings method values the shares at the end of 2028, a later date. A fair comparison puts both on the same day, which the research does by carrying the cash-flow value forward to that date at the cost of equity less the dividends paid on the way.

One way to value a share is to ask how much investors pay for each dollar of annual earnings, the price-to-earnings ratio. On earnings without equity-security gains, Alphabet trades at 32.6x. Applying its own median year-end multiple of 24x to base-case earnings of $13.81 a share for 2028 gives $331.48 at the end of that year. Discounting the cash gives $178.92 today, and $214.41 a share at the end of 2028 on the count that earnings per share uses, which is 15.5x the same earnings. A multiple is not a shortcut that skips the reinvestment. It carries assumptions about how fast profit grows, how much must be reinvested to grow, what the reinvestment earns and how risky the profit is, without showing them. The 24x was observed when the company earned about 50% on its capital and needed little of it to grow, and at our discount rate even a steady state earning that return would be worth about 17.0x earnings. A multiple is a comparison with how the market priced the shares in other years and not independent confirmation of the cash flows.

Uncertainty and sensitivities

A reference path is one illustration and not a forecast. The research tested 33 inputs one at a time and in combinations. Capital spending alone moves the base cash-flow value between $212 on a restrained path and $143 on a continuing one while revenue barely moves, so a precise base value should be read as a point in a range. The research also records tests before the events they concern, with thresholds that come from its own model, and a single quarter usually cannot settle a ten-year return, so the usual honest result is unresolved.

Two more ideas keep estimates honest. A threshold asks what would have to be true, given a hurdle and a set of assumptions: for example, what the capital spent in 2027 would have to earn, beyond the Cloud revenue it carries, to earn a ten percent return, which comes to about $43.8 billion a year before tax on the research's treatment. It is not an estimate of what the capital earns. A ceiling works the same way: an envelope that credits the capital with generous shares of the growth in Search and subscriptions comes to $39.7 billion of contribution in 2028. It is the most that envelope allows, inside assumptions Blockhaven chose, and not a limit on what the capital could do. Neither a threshold nor a ceiling is ever added to a forecast.

A test that stops before the capital's economic life ends leaves out what comes after. If it stops early and counts only the flows inside it, that is the same as setting every later cash flow to zero, which asserts that the equipment earns nothing afterwards. A truncated test says as much about its horizon as about the capital. That is why the valuation runs to 2035 and includes a terminal value, itself an assumption about the later years.

How to compare your view with the Street

A Street target is a price a year ahead on a method and a date that a research house chooses. A value in this research is on a stated date and method. Before comparing them, ask three questions: are the periods the same, are the definitions the same, and are the analysts the same group.

Here the comparison is narrow. Revenue for 2026 and 2027 agrees with the public consensus to within about one percent, so for those years the question is profit and not demand. Earnings per share for 2028 are $13.81 in our base case against $18.34 in the average on Nasdaq's page, captured on October 4, 2026, from 11 analysts, on a basis the page does not state. Before treating the difference as disagreement, ask what each figure leaves out: Alphabet publishes no adjusted earnings per share, so an adjusted figure is its provider's own construction, and an add-back of stock-based compensation alone would be worth about $2.82 a share of our 2028 earnings. The average target of $429.36, as StockAnalysis shows the consensus of S&P Global Market Intelligence (captured on October 4, 2026), is 23.4x the Street's own 2028 earnings, which is a ratio and not a multiple any analyst chose. The public pages have since moved; the figures here are as captured.

Consensus is not the truth, and it is not evidence that analysts ignore capital spending. The public sources share analysts, so their counts are never added, a rating label stays in its own scale and a Strong Buy is not a five out of five, and neither the spread of targets nor the number of analysts is a probability.

Why a management statement is not a measurement

A senior executive at Google Cloud said that servers in aggregate pay back in under 2 years, and that on Alphabet's own chips the period is half that of the usual alternative. The statement carries an asterisk and does not say what is counted, over which servers, or whether the numerator is revenue or profit. Keep three layers apart: what management said, what reported figures show afterward and what Blockhaven believes follows. A payback statement is context, not an input. A disclosed return on invested capital that can be checked against reported depreciation would be worth more than any of the statements, and it is the first thing the research's decision rules ask for.

What makes a thesis stronger, weaker or contradicted

A thesis is stronger when evidence arrives that it said would arrive. For the cautious view here, that would be depreciation rising faster than the revenue it supports, or a Cloud margin that falls back toward its level of a year ago as rented capacity and TPU systems arrive. For the bull case it would be a Cloud margin that holds above its second-quarter level through those changes, or a disclosed return on the new capital that can be checked. A thesis is weaker when the evidence it needs fails to arrive for longer than expected, or when the facts drift toward the other case without yet crossing a line the research drew in advance.

A claim is contradicted when reliable, comparable evidence conflicts with what the claim asserts. A verified return on new capital above its cost would weaken concern about the investment program; it would not by itself overturn a cautious stock rating, because the share price may already reflect attractive returns. One quarter rarely settles a long-term question, and unresolved is a valid outcome. The business can prove stronger than expected while the rating remains cautious.

What would change the thesis

• Does Cloud margin hold after depreciation? A margin that holds as rented capacity and depreciation arrive supports the operating case. It does not by itself show what the marginal capital earns.
• Does the backlog become revenue? Look at cumulative Cloud revenue against the amount the stated conversion implies, not at any one quarter. The backlog balance alone does not say.
• Does Search hold up after the comparison gets hard? Growth, the traffic acquisition rate and the Services margin after the third quarter of 2026 are evidence about resilience and monetization, not proof of what caused the growth.
• What is 2027 spending guided to? More spending than the base path, without corroborating evidence of demand, would weigh on the base case.
• What does the Search appeal decide? A ban on default payments would change the cost of Search distribution, for better or worse.

Before investing, decide what evidence would weaken your thesis. A view that can explain only what would prove it right is not a useful investment framework.

How to read Alphabet from here

Each quarter ask five questions: Is Cloud's margin holding? Is the backlog becoming revenue? Is Search growing against a hard comparison? Is depreciation rising with spending in a way revenue can carry? Is management disclosing a return? Compare the answers with the bear, base and bull cases. Do not rewrite the thesis around one quarter. This guide offers no recommendation.

Short glossary

Read more

• The original article on Substack, published October 21, 2026, which carried both the Research Note and this Companion: Alphabet (GOOGL): A Great Franchise, and a Price That Assumes the New Capital Is Great Too (opens in a new tab)
• The Research Note on this site: Alphabet: The Franchise Is Exceptional; the Price Assumes the New Capital Is Too
• A longer Deep Dive (64 pages) holds the full evidence and calculations. It is a separate product, is not part of this page, and is not freely available here.

Disclosures

This learning edition explains the research in the Alphabet Research Companion published October 21, 2026. Scenario values remain conditional, opinions may change as evidence changes, source records can contain limitations, and no scenario is a promise of performance.

Blockhaven research is for informational purposes only and is not investment advice. Not a recommendation to buy or sell any security.

Important disclosures

Blockhaven research is provided for general informational and educational purposes. It is not individualized investment advice and does not take into account any reader's financial circumstances, investment objectives, risk tolerance, tax situation or portfolio.

Nothing in this report is an offer or a solicitation to buy or sell a security or other financial instrument. Reading or accessing Blockhaven research does not by itself create an investment-advisory, fiduciary, brokerage or client relationship.

Investing involves risk, including the possible loss of principal. Scenario values, terminal values, valuation ranges, sensitivities and forward-looking assumptions are analytical tools, not guarantees or promises of future performance, and actual results may differ materially.

Blockhaven uses information from company filings, company publications, market-data providers and other sources it believes to be reliable, but that information may be incomplete, delayed, revised or incorrect.

Opinions, ratings and interpretations reflect Blockhaven's research judgment as of the report date and may change as new evidence becomes available.

Conflicts and positions

Position Disclosure: As of 10-20-2026, the author holds no shares of Alphabet Inc. (NASDAQ: GOOGL, GOOG). The author holds no options on the company. The author has no paid financial relationship with the company. That includes any paid advisory or consulting relationship. It also includes any paid sponsorship or advertising relationship. This disclosure states no other position or relationship.

What a Blockhaven Rating means, and what it is not: Research disclosures.