The AI build-out is projected to demand substantial capital, with global spending on AI-related data centers and supporting infrastructure potentially reaching $7 trillion by 2030, according to McKinsey. This could be one of the largest peacetime investment projects ever, with five American hyperscalers (Amazon, Alphabet, Meta, Microsoft, and Oracle) expected to spend approximately $800 billion in 2026 and nearly $1.1 trillion in 2027. However, the economics of this technology remain unproven, leading to concerns about over-investment and capital misallocation if AI demand slows, potentially leaving data centers built ahead of need and capacity exceeding demand.
While hyperscalers have historically funded these investments from operating cash flows, they are increasingly relying on debt. In 2025, these five companies issued a record $121 billion in debt, more than four times their average annual borrowing between 2020 and 2024. Morgan Stanley analysts predict global data center spending will hit almost $3 trillion between now and 2029, with only $1.4 trillion covered by Big Tech's capital expenditure, leaving a mammoth $1.5 trillion to be financed by investors and developers. Companies like Meta are already raising significant debt, including $26 billion from private capital investors this month, to fund data center projects.
Key risks include the rapid obsolescence of AI technology, which could require new investments and decrease returns, or force asset sales at a discount. If AI demand plateaus or more efficient training models emerge, vast sums could be tied up in stranded assets. For example, OpenAI had planned to spend $1.4 trillion on data center rentals over eight years but has since moderated this to $600 billion over four, and discontinued its power-hungry video generator Sora, signaling a shift toward financial discipline. The tech giant Apple is notably absent from this data center construction craze, maintaining a lean balance sheet compared to its peers.
