Google is also saying computing power isn’t enough: Alphabet increases its stake by $80 billion to secure AI infrastructure, while Berkshire Hathaway unusually pours more than $10 billion into a share investment.

Alphabet announced a $80 billion stock issuance plan to expand AI data centers and computing infrastructure, with one reason: demand has exceeded supply. Of this, $10 billion comes from Berkshire Hathaway, which will take the position via a private placement.
(Background: Gavin Baker’s barbell strategy as an early investor in Nvidia—long AI infrastructure, short the broad market)
(Additional context: In The Big Short, Michael Burry roars that SpaceX can’t support a trillion-dollar valuation, and bluntly calls Anthropic’s computing “frenzy” a mirage.)

Table of Contents

Toggle

• Where does the $80 billion come from, and how is it sold
• Why does the “company that lacks for money the least” still raise funds externally
• The $700 billion arms race across the whole industry

One of the companies with the most cash in the world has decided to reach out to the market for money. Not because it can’t keep going—but because AI’s appetite is bigger than anyone could imagine.

On Monday, Alphabet announced a stock issuance plan on the scale of $80 billion, and there is only one reason: “demand has exceeded supply.”

Where does the $80 billion come from, and how is it sold

This money will be raised through three channels.

First, $30 billion goes through a public offering: of which $15 billion is mandatory convertible preferred stock. Put plainly, investors buy this “special preferred stock” first, and then, at maturity, it is forcibly converted into common stock. In other words, the company gets the money now, and equity will be diluted in the future; in addition, the other $15 billion is directly issued as A shares and C shares common stock.

Second, $40 billion goes through an ATM issuance, with the plan to begin in Q3 2026. ATM is short for “At-the-Market.” Put simply, it means selling shares on the market at any time in small batches, rather than dumping a large amount all at once—so the impact on the share price is relatively controllable.

Third, $10 billion will come from Berkshire Hathaway via a private placement and direct investment, without going through the public market.

With these three tracks opened at the same time, the underlying logic is to diversify fundraising risk and simultaneously lock in different types of capital sources—institutional investors, the retail market, and long-term strategic shareholders each getting what they need.

Why does the “company that lacks for money the least” still raise funds externally

For years, Alphabet has maintained a large cash balance and short-term investments. Search ads, Google Cloud, and YouTube continue to generate strong free cash flow. The normal logic is: a company like this basically doesn’t need to ask the market for money.

Yet Alphabet has chosen to issue new shares. The message behind it is this: the scale of AI capital expenditures has grown so large that it’s no longer something even its own cash flow is willing to shoulder on its own. Management’s judgment is to diversify part of the funding needs and the corresponding risks into the capital markets, rather than burning through its own ammunition.

Raising more equity rather than taking on debt is a deliberate financial choice. The upside of issuing debt is that it does not dilute shareholders, but the downside is that the company will have to repay principal and pay interest on a regular schedule in the future, creating a rigid cash-flow burden. The cost of issuing equity is “dilution.” In plain terms, if shareholders initially held 1% of the company, after the issuance their ownership percentage will shrink slightly because the denominator (the total number of shares) becomes larger—though the benefit is that there is no fixed repayment obligation, giving the company greater financial flexibility.

Management clearly believes that the payback timeline for this round of AI investment is long. Rather than taking on periodic debt repayments, it makes more sense to use equity to obtain longer-term capital with greater flexibility, so that the balance sheet can remain healthy.

At a recent Google I/O, Sundar Pichai said that the estimated capital expenditures for 2026 will reach $175 billion to $185 billion, revised upward again from $175 billion to $185 billion in April—wait, the original statement indicates upward revision from April, so the correct figures in the source are: 2026 capital expenditures are estimated at $180 billion to $190 billion, up again from $175 billion to $185 billion in April. With $80 billion raised, that is only enough to cover about 45% of annual capital expenditures, and the rest still needs to be filled with its own free cash flow.

The $700 billion arms race across the whole industry

Bloomberg estimates that AI capital expenditures across the whole industry this year will top out at $700 billion. Alphabet’s $80 billion fundraising, added on top of its own projected annual capital expenditures of $180 billion to $190 billion, corresponds to the industry’s $700 billion “pie”—it is only one slice, but it is an extremely heavy one.

In this arms race, it is also worth noting that Berkshire Hathaway has chosen to enter at this moment. Berkshire is known for its value-investing philosophy, and for years it has kept its distance from money-burning technology sectors. Even when it later doubled down on Apple, it acted only after Apple had demonstrated strong cash flow and a durable moat.

Now, Berkshire is making a $10 billion private placement direct bet on Alphabet’s AI infrastructure. This is widely recognized as the most conservative kind of long-term capital—actively placing money into one of the most cash-burning tracks. The contrast itself sends a strong signal: in the view of prudent capital, AI infrastructure is no longer just speculative, but an asset with moat-like characteristics and long-term compounding potential. The specific shareholding structure and terms of this private placement have not been fully disclosed yet, but the direction alone is already enough to make the point.

The flywheel logic in the AI track is: the more computing power there is, the better the models that can be run; the more customers that can be served, the higher the revenue; and then that revenue, in turn, supports the next round of capital expenditures. At present, in the short term, there still does not appear to be a braking point to this logic.