The announcement on June 1 was the largest equity raise Alphabet has ever attempted. Two days later, after the underwritten portion was oversubscribed, the company upsized it. Bloomberg reported the final figure at $84.75 billion — call it $85 billion — a sum larger than the combined valuations of most AI startups, and a sign of how far the company will go to keep its data centers full of chips.
The package has three parts. An underwritten public offering of roughly $30 billion, which drew enough demand that Alphabet priced and allocated about $35 billion. A private placement of $10 billion to Berkshire Hathaway, the only investor that gets to write a check that size without blinking. And a $40 billion at-the-market program, which lets the company sell shares gradually over time rather than dump them in one go. The structure lets Alphabet tell two stories at once: the offering is done, the flexibility is open.
The market’s first reaction was skepticism. Alphabet shares fell about 4% on June 2, the day after the plan was announced, as investors absorbed the prospect of new shares and the dilution that comes with them. Management treated the drop as the price of doing business. Chief Executive Sundar Pichai and Chief Financial Officer Anat Ashkenazi held an investor presentation to explain why the company was selling stock instead of borrowing: the demand for compute, they argued, is running ahead of the company’s ability to supply it, and the gap is already booked.
The backlog number makes the argument concrete. Google Cloud’s contracted demand — revenue under contract but not yet recognized — roughly doubled in a single quarter, from about $240 billion to approximately $460 billion, according to the company’s investor materials. That is not a forecast; it is a pile of signed commitments waiting for capacity. Alphabet raised its 2026 capital expenditure guidance to between $180 billion and $190 billion to build toward that backlog, and even that pace, the company has said, may not be enough.
Alphabet’s choice of equity over debt is worth pausing on. For two decades, the company funded itself almost entirely from its own cash generation, and borrowing to build was a habit of other people’s companies. The AI buildout broke that pattern. But when the financing need grew, Alphabet judged that selling ownership was cleaner than borrowing at this scale. Dilution, in other words, was the explicit trade-off: shareholders own a slightly smaller piece of a company that hopes to be much larger.
The money will go into the physical plant of the AI business: data center construction, GPU clusters and the training runs of Gemini, the model family that anchors Google’s product strategy. Pichai framed the raise in his presentation as a response to demand that is supply-constrained, not a speculative over-build. The distinction matters to investors, because the AI infrastructure cycle has produced two kinds of spending: capacity chasing booked demand, and capacity chasing hope. Alphabet is presenting its spending as the first kind.
The timing is not coincidental. Microsoft and OpenAI have deepened their partnership even as it strains. Anthropic is heading to the public market with nearly a trillion-dollar valuation. OpenAI has begun its own IPO process. Every one of Google’s rivals has found a way to stack capital against the same infrastructure. Alphabet, which once could afford to watch the race from the sidelines, decided this month that it could not.
The size of the raise also changes the geometry of the competition. A financing package larger than the entire valuation of most AI companies gives Google a form of endurance that pure startups cannot match: it can keep buying chips, keep training models, and keep building data centers through multiple rounds of market turbulence, because its balance sheet — now supplemented by $85 billion in fresh equity and Berkshire Hathaway as a shareholder — does not have a use-by date.
What investors will watch next is execution. Alphabet’s capital plan now assumes the backlog is real and the capacity arrives on time. The $460 billion figure, if it holds, means the spending is already matched by customer commitments. The risks are the ones every hyperscaler faces: GPU delivery slips, construction delays, and the possibility that model improvements make today’s capacity obsolete before it is fully depreciated.
For Google’s rivals, the raise changes the arithmetic of the race: a company that can raise $85 billion in a week does not have to choose between models, data centers and chips. It can build all three at once, and it has said it intends to. For now, the message from Mountain View is unambiguous. The company that built the modern internet mostly out of its own cash flow has decided that the AI buildout is too big to wait. It sold a stake in itself to buy the future, and the market’s job, from here, is to decide whether that future is worth what Alphabet paid for it.
