What it does · how the money flows · why it matters
Charlie Munger once said missing this company was one of the great mistakes of our careers — that we sat there "sucking our thumbs" while our own insurer, GEICO, paid Google ten dollars a click for something that cost Google almost nothing. So let me start with humility. Alphabet is a toll-booth on the world's information and advertising. When two billion people go looking for something, Google shows them answers — and, alongside, an advertisement that a business paid to place there. That auction, running billions of times a day at almost no marginal cost, is one of the greatest money-machines ever built.
Founded in 1998 by two Stanford students, Google organised the internet so well that its name became a verb in dozens of languages. Today, as Alphabet, it is far more than search: it owns YouTube (the world's largest video platform), Android and Chrome (which carry its search to billions of devices), a fast-growing and now-profitable Cloud business, a world-class AI lab in DeepMind, and Waymo, the clear leader in self-driving cars. In the year ended December 2025 it booked $403 billion in revenue and $132 billion in profit, and it is worth about $4.32 trillion.
"You could see in our own operations how well that Google advertising was working. And we just sat there sucking our thumbs. So we're ashamed. We atone." — Charlie Munger, on missing Google
The reason Munger was ashamed is the reason this is a franchise: an advertising auction has almost no marginal cost. When Google shows one more ad, it costs the company essentially nothing, so nearly every extra dollar of ad revenue falls straight to profit. That is why Google earns operating margins above 30% at a scale of hundreds of billions — and why, for twenty years, it has been closer to a royalty on human curiosity than an ordinary business.
| Founded | 1998 · Mountain View, California (USA) · IPO 2004 |
| Sector / Industry | Communication Services · Internet Content & Information |
| CEO | Sundar Pichai (Alphabet CEO since 2019) |
| Makes money from | Advertising (~73%), Cloud (~15%), subscriptions/devices |
| Revenue (FY2025 · TTM) | $403 B · ~$423 B |
| Market capitalisation | ~$4.32 T · newly a Dow-30 member |
How a search box became one of the world's great franchises
The road matters, because it shows both a company of extraordinary technical vision and one that has, more than once, been caught flat-footed and recovered — most recently by artificial intelligence.
| Year | Milestone |
|---|---|
| 1998 | Larry Page & Sergey Brin found Google — 'PageRank' ranks pages by links, not keywords. A better search is born. |
| 2000–03 | AdWords + the pay-per-click auction (bid × quality) — the economic engine of the whole company. AdSense extends it to the web. |
| 2004 | IPO by Dutch auction. Dual-class shares from day one lock in founder control. |
| 2005–08 | Buys Android (~$50M — a legendary deal) and YouTube (~$1.65B); launches Chrome. The distribution empire is built. |
| 2014 | Buys DeepMind. With Google Brain, invents the Transformer (2017) — the 'T' in GPT — a decade before the LLM boom. |
| 2015 | Reorganises into Alphabet; splits the cash machine (Google) from the moonshots (Other Bets). Sundar Pichai runs Google. |
| 2022–23 | ChatGPT triggers an internal 'code red.' Bard's flubbed demo wipes ~$100B off the market cap in a day. |
| 2024–26 | Gemini reaches frontier parity; AI Overviews reach ~2B users; Cloud turns profitable; Alphabet joins the Dow. |
Two lessons for an owner. First, the capital allocation is among the best in history: $50 million for Android and $1.65 billion for YouTube may be the two finest acquisitions ever made. Second — and this is the caution — Google was genuinely blindsided by ChatGPT, and its first answer was a public embarrassment. It recovered impressively (Gemini is now a frontier model). But the episode proved that even this fortress can be surprised, and the surprise came at the one place it could least afford: the front door of search.
How it works — and how confidently you can know its future
The what is beautifully simple. The ten-year picture is, for the first time in two decades, genuinely uncertain. Here is the machine:
This is the honest heart of the matter. For twenty years Google's future was as predictable as Coca-Cola's — you could see the search box being there in a decade. Generative AI has, for the first time, put a real question mark over how people will find information and how well that will pay. The business is understandable; its next chapter is not yet written. That is precisely the tension a careful owner must price.
Revenue by line · geography · the ad auction
Three quarters of Alphabet is still advertising, but the shape is changing — Cloud is now a real second engine, and subscriptions are a fast-growing third. FY2025 revenue by line:
How the auction works. Every time Google shows ads, it runs a real-time auction: advertisers bid a maximum per click, and Google ranks them by bid × quality (how relevant and likely-to-be-clicked the ad is), with the winner usually paying just enough to beat the next bidder. Because showing one more ad costs Google almost nothing, this is a business of extraordinary operating leverage — Google Services earns ~42% operating margins, and those fat profits pay for Cloud's build-out, Waymo's losses, and the enormous AI capex. The whole edifice rests on that one, magnificent, near-zero-marginal-cost auction.
Wide and deep — but, for the first time, under attack
Google's moat is one of the widest ever built. It is filled from five springs:
Scale & data network effects. More searches → more data → better results and targeting → more users and advertisers → more data. This flywheel has held Google near ~90% of search for twenty years.
A two-sided marketplace. Billions of users on one side, millions of advertisers on the other — neither can easily leave the other.
Distribution. Google pays ~$20 billion a year to be the default search in Apple's Safari, and owns Chrome (~65% of browsers) and Android (~70% of phones). Most people never change a default — so owning the default is worth a fortune.
Brand. "To google" is the verb for the whole category — mindshare no rival can buy.
YouTube & Cloud stickiness. A creator-viewer network no one can rebuild, plus real switching costs in Workspace and Cloud.
The honest caveat. The distribution, data and brand springs are structurally durable. But the product itself — the ten-blue-links page — is, for the first time, under genuine attack from AI answer-engines. So I score the moat an 8, not a 10: still vast, but contested at the front door in a way it has never been before.
The single most important issue in the whole analysis
Everything turns on this. When an AI chatbot simply answers your question, you may never see — or click — the ten blue links and the ads beside them. Does that break Google's money-machine, or does Google simply become the best answer-engine and keep the tolls? Here are both sides, honestly.
| The bear case | The bull case |
|---|---|
| AI answers replace the links — and the ads beside them | Google is a top-tier AI lab — Gemini reached frontier parity; it invented the Transformer |
| Fewer searches per task — searches/user reportedly ~-20% | AI Overviews reach ~2B users; management says they monetise at ~parity with search |
| One AI answer shows far less ad inventory than a results page | Custom TPU chips let Google serve AI cheaper than rivals renting Nvidia |
| Innovator's dilemma — cannibalising a 90%-margin cash cow | Unmatched distribution — Chrome/Android push Gemini to billions instantly |
The single most important data point, if it holds, is Google's claim that AI Overviews monetise at roughly the same rate as classic search. If that is true, the franchise survives the transition with its distribution intact — Google simply changes the shape of the answer and keeps collecting the toll. If instead AI answers structurally monetise worse — one paragraph where there used to be a page of high-intent ads — then Google could "win" the technology and still lose margin and revenue for years. That is the irreducible judgment call, and no one — not management, not I — can yet prove it either way. It is the reason this business, uniquely among the great franchises, cannot be bought with total confidence today.
Search & AI · advertising · cloud · autonomy
| Arena | Who | Where Google stands |
|---|---|---|
| Search / AI | OpenAI/ChatGPT, MS Copilot, Perplexity, Anthropic | #1 but contested — Gemini now at parity |
| Advertising | Meta (social), Amazon (retail media) | Co-leads the duopoly; Amazon nibbling |
| Cloud | AWS #1, Azure #2 — Google #3 | #3 but fastest-growing (~48%), now profitable |
| Video / attention | TikTok, Netflix, Meta Reels | YouTube dominant, esp. on the TV |
| Autonomy | Tesla robotaxi (Cruise exited) | Waymo the clear leader — ~$126B value |
Read that table the right way. In its core — search and advertising — Google faces the most serious competition in its history, but from a position of overwhelming strength. In cloud, it is only #3, but it is the fastest-growing of the three and has finally turned profitable, with its AI stack (Gemini + TPUs + DeepMind) as the differentiator. And in Waymo it holds a genuinely commanding lead in what could be an enormous future market — an option most owners get almost for free inside the $4T mothership. Few companies fight on so many fronts; fewer still are winning or gaining on nearly all of them.
Who runs it · who really controls it
Able managers — and a governance structure every investor must understand before buying a single share.
A Buffett-style investor reads the dual-class structure two ways. On the one hand, it lets the founders make very long-horizon, unpopular bets — Waymo, the vast AI capex — without being second-guessed by the market each quarter, and their record earns real trust. On the other, public shareholders have essentially no governance recourse if that judgment ever goes wrong. You are, in plain terms, a passenger — riding with owner-operators who have been superb, but a passenger nonetheless. Know that before you board.
A gusher of cash · and a capex bill that hides it
| Metric | Value | Read |
|---|---|---|
| Revenue growth (5-yr CAGR) | ~17% | ▲ fast for the size |
| Net income (FY2020 → FY2025) | $40B → $132B | ▲ ~3.3× |
| Gross margin | ~60% | ▲ high |
| Operating margin | ~32% | ▲ Services alone ~42% |
| Return on equity (ROE) | 39% | ▲ excellent |
| Return on invested capital (ROIC) | 19% | ▲ well above cost of capital |
| Balance sheet | Net cash · Altman-Z 13.9 | ▲ fortress |
| Capex / revenue | ~26% | ▲ $91B FY25 → ~$185B guided FY26 |
| Free cash flow / share vs EPS | $5.33 vs $13.24 | ◆ FCF crushed by the AI build |
The core is a money-gusher — 39% on equity, near-zero debt, tens of billions in profit a quarter. But note the same asterisk as Microsoft, only bigger: free cash flow ($5.33/share) is a fraction of reported earnings ($13.24), because Alphabet is pouring an astonishing ~$185 billion into AI data centres this year — and Mr. Pichai says even that "won't be enough." Almost none of that is maintenance; it is a colossal growth bet. Whether it earns an adequate return on so vast a capital base is the second great unknown here — and it is why the free-cash-flow figures, and the models built on them, are so badly distorted right now (Part X).
Cheap on earnings, 'expensive' on cash — and why the DCF lies
| Yardstick | Today | Forward | Read |
|---|---|---|---|
| P/E — reported earnings | ~27x | ~25x (FY26) · ~21x (FY28) | fair for the quality & growth |
| P / Free cash flow | 67x | — | distorted — capex-crushed |
| EV / EBITDA | 20x | — | reasonable |
| Price / Sales | 10.2x | — | full but not extreme |
| PEG (trailing) | 0.59 | — | growth-adjusted, attractive |
Here is a genuinely instructive mess. A naive discounted-cash-flow model spits out an intrinsic value of about $132 — implying the stock is 63% overvalued. Ignore it. That model treats today's $185 billion of growth capex as a permanent expense, crushing free cash flow and the value built on it — it is artificially, absurdly low. Wall Street's analysts, looking through the capex to earnings, land near $413 (+16%). The truth sits between, and closer to the analysts: on ~27× earnings — even ~25× next year's — for a wide-moat franchise still growing in the mid-teens, with Cloud inflecting and Waymo thrown in, Google is fairly-to-attractively priced. This is the mirror image of the free-cash-flow trap: the cheapest way to misjudge Alphabet today is to trust its free cash flow.
The courts · the AI question · the capex bet
The two biggest risks are the two open questions: does AI-answer search monetise like classic search (Part VI), and will the $185B capex earn its keep (Part IX). On antitrust, the news is better than the headlines: in the DOJ search case a judge ruled Google an illegal monopolist (2024) but in September 2025 declined to break off Chrome or Android — ordering instead an end to exclusive defaults and some data-sharing (both under appeal into 2026). A separate ad-tech case and ~€11B of EU fines sting the cash flow but do not threaten the franchise. The honest summary: the existential legal risk has receded; the AI and capex questions have not.
Let me begin with a confession, because it frames everything that follows. Charlie and I missed this company — badly — and we knew better. Our own insurer was paying Google handsomely for advertising that cost Google almost nothing, and we sat on our hands while a great franchise compounded in plain sight. I have carried that lesson for twenty years, and I bring it to the page today so that I do not make the opposite error — dismissing a wonderful business because it now wears a fashionable worry.
Because a wonderful business is exactly what this is. Alphabet is a toll-booth on the world's information and advertising, with a search franchise that has held ninety percent share for two decades, the largest video platform ever built, a browser and a phone operating system that carry its products to billions, a cloud business that has finally turned profitable and grows near fifty percent a year, one of the two or three best AI laboratories on Earth, and — almost as a free option — the clear leader in self-driving cars. It earns thirty-nine percent on equity, carries a fortress balance sheet, and is run by owner-operators with a capital-allocation record — Android for fifty million dollars, YouTube for a billion and a half — that belongs in a textbook. On quality, this is a franchise of the very first rank.
And unlike the chip-maker we looked at recently, I can broadly understand where it earns its keep: an advertising auction is no harder to grasp than a lemonade stand, and it is one of the finest business models ever devised. Nor is it priced for the heavens — at about twenty-seven times earnings, even twenty-five times next year's, this trades like an ordinary blue chip, not a mania. I would warn you, in passing, to ignore the discounted-cash-flow model that screams the stock is worth a third of its price: it is fooled by the enormous, temporary bulge of artificial-intelligence spending, treating a growth bet as if it were a permanent cost. On the earnings a franchise like this reliably produces, the price is fair — and toward three hundred dollars, genuinely attractive.
So why do I stop short of pounding the table? Because for the first time in twenty years, the front door of the castle is contested. When a person asks a machine a question and simply receives an answer, they may never see the ten blue links — or the advertisements beside them — that have paid for all of this. Google's own claim is that its AI answers monetise about as well as its old results page; if that is true, the franchise sails through, and this letter will read as far too cautious. But if one synthesised paragraph structurally earns less than a page of high-intent ads, then Google may win the technology and still lose margin for years — the classic dilemma of cannibalising your own cash cow. I cannot yet prove which way it breaks, and neither can anyone else. Add a hundred-and-eighty-five-billion-dollar annual capital bill of unproven return, and a control structure that makes you a passenger rather than a director, and you have real, unresolved questions sitting beneath a magnificent business.
What, then, would I do? I would own it — watchfully. This is not the chip-maker in the "too hard" pile; it is a franchise I understand, at a fair price, with two honest question marks I can monitor quarter by quarter — search revenue per query, and the returns on that capex. I would take a position here and let it compound, size it for the uncertainty rather than bet the farm, and add with real conviction on any weakness toward three hundred dollars, where the margin of safety opens and Mr. Market hands me the doubt at a discount. Having once sat sucking my thumb while this very company ran away from me, I do not intend to make the same mistake twice — but nor will I pretend the question at its front door isn't real.