Google’s AI Data-Center Spending Outruns Its Cash Flow

Alphabet’s (GOOGL ) Google Cloud unit booked $24.8 billion in second-quarter revenue, up 82% from a year earlier, as enterprises and AI labs rented more of Google’s compute than the company could build. The second-quarter results, reported July 22, 2026, lifted total Alphabet revenue 24% to $119.8 billion and pushed cloud growth well past the 63% rate of the first quarter. They also exposed what that growth costs: Alphabet spent $44.9 billion on property and equipment during the quarter, more cash than the business generated, and its free cash flow turned negative for the first time in years.

Free cash flow, the cash left after capital spending, came in at negative $5.9 billion, down from positive $10.1 billion in the first quarter of 2026. Operating cash flow itself was healthy at $39.1 billion; the deficit came entirely from capital expenditure that roughly doubled from $22.4 billion a year earlier, most of it servers, AI accelerators, and data centers. Alphabet also stopped repurchasing its own stock, spending nothing against $13.2 billion in the same quarter of 2025, a reversal for a company that spent the past decade returning cash to shareholders.

What the cloud numbers show

For now, the spending is converting into a profitable business. Google Cloud’s operating income more than tripled to $8.8 billion from $2.8 billion, a sign the capacity Google is building is being rented at a margin rather than sold below cost to win share. Google Cloud remains the third-largest US cloud platform behind Amazon (AMZN ) Web Services and Microsoft Azure, but it has been expanding far faster than either. CEO Sundar Pichai tied the acceleration to “demand for AI infrastructure and AI solutions,” and said nearly 90% of the Fortune 100 now use its Gemini Enterprise platform.

The demand shows up in raw consumption: Alphabet said Gemini models now process 22 billion API tokens a minute and its Gemini app has 950 million monthly users. Serving that traffic increasingly runs on Google’s own silicon. The company has begun disclosing that Google Cloud’s product revenue comes chiefly from selling TPU systems, its custom AI chips, and in June it said it would ship those chips to select customers to install in their own data centers, opening a market it had kept inside its own facilities. Pichai has told investors that demand is running ahead of the compute Google can bring online, the same squeeze pushing rivals toward deals like Meta’s talks to rent capacity from Anthropic.

The funding shift

Paying for the build is forcing Alphabet to raise outside money, something it rarely needed to do. In June 2026 it issued $49.6 billion in stock and mandatory convertible preferred shares, earmarking the proceeds to scale AI infrastructure and global compute, and arranged to sell up to $40 billion more. It issued $20.3 billion in senior notes during the quarter, lifting long-term debt to $98.2 billion from $46.5 billion at the end of 2025. The turn toward debt markets mirrors a wider shift on Wall Street, where AI infrastructure has become a major engine of new borrowing.

The capital plan is still climbing. Alphabet raised its capital-spending guidance for 2026 to $195 billion to $205 billion on the same report, up from the $180 billion to $190 billion range it gave in April, roughly six times its 2022 level, and told investors 2027 will be higher again. The increase, not the quarter’s results, is what sent the stock down about 5 percent after hours. Property and equipment on its balance sheet has already swelled to $321 billion, and rising depreciation on that hardware, which management has flagged as a drag on margins, will press on profits as the assets come online. The build also competes for the scarce inputs — power, land, and chips — that have turned data-center capacity into AI’s real bottleneck.

Reading past the headline profit

Alphabet’s headline profit looked enormous, with net income of $112.1 billion, up nearly 300%, but almost none of it came from the operating business. A $99 billion unrealized gain on equity holdings, mostly its stakes in SpaceX (SPCX ) and Anthropic, drove the number; operating income, the cleaner read, rose 30% to $40.8 billion. For anyone tracking the economics of the AI build, the more telling lines sit lower in the statement: the $44.9 billion of capital spending and the negative free cash flow beneath the record profit.

The quarter captures the bet every hyperscaler is now placing. Cloud revenue up 82% and operating income tripling say demand for AI compute is real and monetizable today; the negative free cash flow says Google is spending faster than that demand pays it back, wagering the gap closes before depreciation catches up. Whether that math holds is the question hanging over the entire infrastructure boom.