“I feel like a horse’s ass for not identifying Google. We screwed up.”
— Charlie Munger
Listen to the deep-dive discussion – Why Berkshire Hathaway Bet on Alphabet AI
Download the Slide Deck (PDF) The_Alphabet_Architecture
I. The Decision Worth Defending
When Berkshire Hathaway committed $10 billion to Alphabet at roughly $350 per share, it did so on the same day Alphabet announced it would sell $80 billion of stock and spend $190 billion on capital expenditure. A previous post used those numbers to argue that Google revealed an Imperfect B—growth that costs money, deployed outside the moat that protects Search. That argument has merit. This post makes the opposite case, because the opposite case is strong, and because Berkshire’s decision deserves to be understood as the intelligent bet it almost certainly is.
First, a brief reminder of the framework’s vocabulary. Every business has an A—the toll it collects today—and a B—the wave of growth that compounds the toll over time. When the B arrives inside the moat at zero cost, the business is a Freesurfer, and its growth is free. When the B requires capital, effort, and competition outside the moat, it is mechanical—an Imperfect B. Google’s Search advertising is a natural B inside a near-perfect moat. Its AI and cloud ambitions are the mechanical kind. Hold that distinction; the entire case turns on it.
The case for Google rests on a principle Buffett himself articulated—one that complicates the Freesurfer framework rather than confirming it. The best business to own, he said, is one that over an extended period can employ large amounts of incremental capital at very high rates of return. The rarest and most valuable business is not the one that needs no capital. It is the one that can absorb large and growing amounts of it while still earning extraordinary returns on every dollar.
By that standard, the $190 billion is not a weakness. It may be the entire thesis.
II. The Limitation of the Capital-Light Business
Buffett loves See’s Candies. It earns extraordinary returns on the capital it employs. But he has been candid about its limitation: you cannot deploy meaningful new capital into See’s at those returns. The candy market is saturated. The business throws off cash that must be invested elsewhere, because See’s itself cannot absorb it.
Coca-Cola has the same constraint. Magnificent returns, but a slow wave and limited capacity to reinvest. The cash comes out. The shareholder must find a new home for it.
This is the hidden cost of the capital-light business—the cost the Freesurfer framework rarely names. A business that returns most of its cash creates a reinvestment problem. Every dollar returned through a dividend or a buyback is a dollar the shareholder must redeploy at attractive rates, in a market that may not offer them. The capital-light compounder is wonderful—but it hands the hardest part of compounding back to you.
III. The Business That Solves Reinvestment
The ideal business eliminates this problem. It earns high returns and can absorb a growing pile of capital at those same returns, year after year. The shareholder never has to find a new home for the cash. The business itself is the home.
These are extraordinarily rare. Most businesses fall into one of two disappointing categories: they earn high returns but cannot absorb capital (See’s), or they absorb enormous capital at mediocre returns (airlines, telecoms, utilities). The business that does both—high returns on a large and growing base—is the rarest architecture in capitalism, and it is precisely the one Buffett named as the best business to own.
The case for Alphabet is that it might be exactly this. Google generates more cash than almost any company in history. And unlike See’s, it has somewhere to put that cash—AI infrastructure, cloud computing, the compute layer of the next computing paradigm. If those investments earn high returns, Google is compounding an ever-growing capital base internally, at scale, for decades. That is more valuable than a business that must return its cash because it has nowhere to deploy it.
IV. Why Absolute Returns Matter
There is a second dimension the percentage-obsessed investor misses. Returns outside the moat may be lower than Visa’s—but on a vastly larger base, the absolute dollars can dwarf anything a capital-light business produces.
A 30% return on a small capital base produces a few billion dollars. A 13% return on $190 billion produces nearly $25 billion—every year, growing. The capital-light business runs out of room. The capital-absorbing business at a respectable return keeps compounding a larger and larger sum. In absolute terms—the terms that actually build fortunes—the larger base at the lower rate can win.
This is why Berkshire’s $10 billion in Alphabet may not be a bet on artificial intelligence at all. It may be a bet on a business capable of redeploying an ever-growing capital base at high rates—the best business to own, in Buffett’s own words, financed by a Search moat strong enough to fund the experiment.
V. The Condition Beneath the Principle
The case is strong. But Buffett’s principle carries a condition inside it. The phrase is “very high rates of return”—and the entire bet depends on whether those high rates can be sustained on each new layer of capital, not merely earned on the first.
Sustained high returns on deployed capital require a moat protecting the deployment. When a business reinvests inside its franchise—a proven format expanding, a regulated infrastructure adding to its base—the high returns endure because no competitor can compete them away. That is the best business to own in its fullest form: capital absorbed and protected.
Google’s $190 billion is deployed outside the Search moat, on a level playing field against Microsoft Azure and Amazon Web Services—two competitors of comparable scale and resources. The Search moat does not protect the cloud. So the question is not whether Google can deploy capital. It plainly can. The question is whether the high rate of return will endure against entrenched competition, or be competed down toward the cost of capital, as happens on every level playing field eventually.
VI. Why the Bet Is Intelligent Anyway
Here is why Berkshire is right regardless. Search—the natural B—is so dominant, so deeply embedded in human behavior, so impossible to replicate, that it generates enough cash to fund the entire AI deployment while still leaving the core economics intact. Berkshire is not risking the franchise. It is buying the franchise and receiving the AI deployment as an embedded option—one that may pay off enormously if Google’s scale, data, talent, and distribution convert the level playing field into a genuine edge.
At roughly $350, with a modest discount to market, Berkshire pays a fair price for the most powerful advertising toll in history and receives the optionality on AI infrastructure for what is effectively a reasonable premium. If the deployment earns high sustained returns, Google becomes the best business to own. If it does not, the Search moat still anchors the investment. Heads, extraordinary. Tails, merely very good.
That asymmetry is why the decision is intelligent. It is also why it differs in kind from owning a Freesurfer. The Freesurfer is a bet on architecture under near-certainty—capped, free, and sure. Google is a bet on capital deployment under uncertainty—uncapped, costly, and probable. Berkshire, holding both kinds of businesses, took the probable bet at a fair price, without selling a single share of the certain ones to do it.
The hierarchy is preserved. The Freesurfer remains the purest architecture. But the best business to own, if it can be found, is the one that compounds a growing base at high rates—and Berkshire is betting, intelligently, that Google might become it.
Watch the video
Read the companion post:
$190 Billion to Grow · The Duration Spectrum · The Imperfect B · Toll Position
Based on the framework from The Infinite Investor, available at averagingup.com.