Five Cloud Giants’ 2026 Spending Could Exceed Cash Flow by $93.4 Billion

Five major cloud companies are projected to spend $800.5 billion in 2026, $93.4 billion more than their estimated operating cash flow, sharpening questions about how the U.S. AI buildout will be financed.

Oracle, Microsoft, Amazon, Meta and Alphabet are expected to post combined capital expenditures of $800.5 billion, compared with estimated combined operating cash flow of $707.1 billion. It would be the first year their projected spending exceeds that cash flow.

The figures cover the companies’ overall spending, including non-AI capital expenditures, and exclude leased data centers. They combine reported results with company guidance and analyst estimates, so they do not measure AI spending already incurred.

A longer-term U.S. buildout scenario puts cumulative investment at $10.3 trillion from 2025 through 2032. The modeled total covers data centers, power systems, networking and specialized computing equipment.

The scenario uses a July 2026 inventory of 509 gigawatts of planned capacity and assumes 182.8 gigawatts comes online by 2032. The full pipeline is not a completion forecast: projects may be delayed, reduced or canceled. The estimate also includes spending on some projects scheduled to finish after 2032. A separate $7.6 trillion scenario for 2026 through 2031 used a different time frame.

Leases, joint ventures, project debt, private credit and other arrangements are playing a larger role in financing the expansion. They can bring in outside capital and spread ownership and obligations among investors. Returns remain tied to demand for computing capacity, power access and the financial health of a relatively small tenant group. Grid capacity, transmission access and permits can take years to secure.

At a 50% operating cash-flow margin, the model requires roughly $3.7 trillion in annual revenue at maturity from capacity completed during 2025–2032 to achieve a 10% unlevered return. Different return and margin assumptions put the required revenue between $2.29 trillion and $6.72 trillion. These figures are modeled requirements, not sales forecasts.

If AI investment slows sooner than expected, asset prices could fall, and borrowing and concentrated exposures could deepen the decline. Stijn Van Nieuwerburgh, a Columbia Business School real-estate professor, cautioned, “It would be premature to conclude that AI infrastructure already poses systemic risk comparable to earlier credit booms.” The test is whether demand, profitability and financing can keep pace with planned capacity.