Alphabet just posted its fastest growth in a decade — revenue +24%, Cloud +82%, a $514B backlog — and free cash flow went negative for the first time in company history. The most profitable information business ever built is being rebuilt as an industrial one, at roughly $200 billion a year. Five analyst lenses, three scenarios, four horizons.
The organizing metaphor of this report: what changes when the unit of output stops being a link and becomes an answer.
Gray line = Alphabet Class C’s actual close from the April 2025 trough ($143) through the May 18, 2026 record ($404), the post-earnings flush to $318 on July 23, and back to $343.34 on August 25, 2026. Colored paths = synthesized scenario midpoints forward, probability-weighted 30% bear / 45% base / 25% bull. Log scale, so equal vertical distance means equal percentage move; the left panel is 16 months of actual price and the right panel is five forward years, so the two halves run at different time densities — the dashed TODAY line marks the join. A handful of late-2025 waypoints are derived from Class A closes where Class C prints were unavailable.
Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of every scenario card update live.
The same set of facts — 24% growth, negative free cash flow, a suspended buyback and a 17× trailing multiple — supports wildly different conclusions depending on which framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target and the multiple behind it.
Cloud is no longer a third business — it is becoming a chip business. Revenue accelerated to +82% with a $514B backlog, over half of which converts inside 24 months, and Alphabet recognised external TPU system revenue for the first time in Q2 with the bulk landing in 2027. Gemini has 950M monthly users and AI Mode in Search crossed 1B. The market is still pricing an ad company.
Operating cash flow is $186B and ROIC is 24.9% against a ~10.8% cost of capital — the machine still works. But free cash flow fell 20% to $53B, the FCF yield is 1.26%, the buyback is suspended and net debt swung by $91B in eighteen months. You are being asked to pay a quality multiple for a business whose cash conversion is temporarily unknowable.
Roughly $200B of 2026 capex on four-to-six-year lives is a depreciation wall arriving in 2027–2029, exactly when the AI revenue curve must prove it is durable rather than a land grab. D&A is already $25.5B and rising, interest expense is up nearly five-fold, and the $99B that made Q2 EPS look spectacular was a mark on SpaceX and Anthropic, not operations. Then the frontier team walked out.
Nobody else owns the model, the chip, the data centre, the distribution surface and the ad auction. Vertical integration is worth more, not less, when compute is the scarce input: Alphabet designs its own TPUs (dual-sourced across Broadcom and Marvell, with rights to buy up to $12.2B of Marvell stock), runs them in its own halls, and sells the output through a search box ~89% of the planet already uses.
Alphabet has re-rated from an asset-light compounder into a levered infrastructure builder at the most reflexive point of the AI capex cycle. It raised roughly $70B of external capital in a single quarter — equity, mandatory converts, senior notes, a $40B at-the-market programme and a $10B private placement to Berkshire. That is not what a self-funding franchise looks like, and it imports data-centre-cycle beta the stock never had.
Panel average ≈ $365 — deliberately below the Street’s $422, because three of the five lenses assign real weight to the depreciation and financing risk the sell-side largely models away. These are synthesized frameworks, not real analysts or real firm ratings.
What the sell-side expects over the next year. Bars are sorted low to high and coloured by rating; the dashed line is today’s $343.34. Note what happened around the July 22 print: several desks cut targets while keeping Buy ratings — the capex raise moved numbers, not conviction.
Eight named desks out of the 63 firms covering the stock, all published between July 23 and August 3, 2026. Every one sits above today’s price — but look at the direction of travel rather than the level: J.P. Morgan cut $460 to $420, Oppenheimer $445 to $400, Raymond James $425 to $400 and UBS $400 to $379, all on the same day, all while keeping their ratings. Bank of America went the other way, lifting $370 to $430 on August 3. Morgan Stanley’s $400 (Overweight, August 25) is published against Class A; the two share classes trade within about 1% of each other. Consensus, range and rating counts from S&P Global via TipRanks, August 25, 2026.
Synthesized scenario midpoints, dated mid-year. Returns are versus today’s $343.34. These are illustrative frameworks for thinking about a range of outcomes — not predictions. The clay row in each card is wired to the sliders above.
Where the money actually goes. Both theses live in the gap between the sky bar and the clay bar — and in what happened to the olive one.
Read it left to right. Revenue (sky) compounds from $307B to a consensus $498B — that part of the story is untouched. Capex (clay) goes $32B → $53B → $91B → roughly $200B, a six-fold rise in three years, and management has said 2027 will rise significantly again. Free cash flow (olive) sat flat near $70B for three years and then collapses to a consensus $9.7B in 2026 — the Q2 quarter itself printed −$5.9B, the first negative FCF quarter in Alphabet’s history. Total debt (slate) quadruples from $30B to $121B. FY2026 capex is the midpoint of the $195–205B guidance given on the July 22 call; note the company’s own February guidance was $175–185B, so this figure has moved twice this year. Actuals from the FY2023–25 statements via S&P Global; debt bar is the June 30, 2026 balance.
Every price target on this page is an earnings estimate times an exit multiple. Here is the earnings line — and the reason you cannot use it straight out of the box.
Gray = reported, olive = consensus estimate, clay dashed = the desk’s derived core figure. The 2026E bar of $20.59 and the 2027E bar of $14.81 are both real consensus numbers, and the apparent 28% "decline" between them is an accounting artefact: 2026 contains roughly $99B of unrealised gains on private stakes that do not repeat. Alphabet’s core earning power in 2026 is closer to $11.40 — derived two ways in the collapsible above, and corroborated exactly by the $2.85 adjusted EPS the company reported for Q2. Use the clay line, not the olive one, when you multiply by an exit multiple.
Q2 FY26, year-over-year. Read these against a stock sitting 15% below its May high — that disconnect is the bull’s entire argument in one chart.
Every line except the legacy Network business is growing, and the frontier lines (clay) are compounding off bases that are no longer small: Cloud is a $24.8B quarter, and its operating margin went from 20.7% to 35.6% in a year, tripling segment profit. Search grew 17% — faster than it did before AI Overviews existed, which is the single most important data point against the disintermediation thesis. Alphabet reported its twelfth consecutive quarter of double-digit revenue growth. Cloud operating income growth is derived from the reported margin expansion on reported segment revenue.
The whole valuation argument compresses into one disagreement: is $200B a year the entry fee for owning the next computing platform, or a permanent tax on the best business ever built?
Ten risks placed over a three-to-five-year horizon. Cells heat by severity: the hottest corner is likely × high. The tail row is where the low-probability, high-consequence outcomes sit — the ones that do not show up in anyone’s model until they do.
What breaks: $200B+ of 2026 capex on four-to-six-year lives becomes $40–50B a year of incremental D&A from 2027, compressing operating margin even if revenue keeps compounding at 20%.
What breaks: a US or EU order to divest AdX or the publisher ad server. Behavioural fixes are absorbable; a forced sale removes a profit pool and hands the auction to rivals.
What breaks: the $514B backlog slips past the guided 24 months, or is concentrated in a handful of AI labs whose own funding turns. Capacity built for it becomes stranded depreciation.
What breaks: a generated answer costs orders of magnitude more compute than a link and carries less ad inventory. Revenue per query can grow while gross margin per query falls.
What breaks: management already flagged Q3 pressure from renting third-party capacity, and TPU hardware sales carry roughly 30% gross margin against a 35.6% segment operating margin built on software.
What breaks: Dean, Vinyals, Le and Ghemawat left together in August 2026 and Hassabis stepped back from operations. Frontier model leadership is people-dependent in a way ad auctions never were.
What breaks: the buyback is suspended, ~$70B was raised in one quarter, and a $40B at-the-market programme sits open. Share count rising while EPS is under pressure is a double hit.
What breaks: a rival frontier model plus an agent surface that transacts on the user’s behalf makes the search box a background API. Low odds on today’s data — Search still grew 17% — but it is the only risk that ends the franchise.
What breaks: the ~$99B of unrealised gains on private stakes reverses on a down round or a broad AI repricing, hitting book value and turning a GAAP EPS tailwind into a headline loss.
What breaks: nothing structural — €4.1B Android (now final), €2.95B ad-tech and €890M DMA are each under a quarter of operating income. The cost is compliance drag and follow-on damages, not solvency.
Hover the dotted terms in the metrics strip above, or scan the desk’s working definitions here.