Hyperscale cloud providers are expected to spend nearly $1.1 trillion on AI data centers by 2027, according to University of Pennsylvania finance professor Jessica Wachter. To justify this massive investment, these companies must increase their productivity by 2.7 times by 2030, factoring in capital costs, asset depreciation, and a 15% return requirement, Wachter said in an analysis published by technologyreview.com.
Wachter approached the AI infrastructure boom with a straightforward accounting framework, focusing on the earnings growth needed to break even on the hyperscalers’ expenditures. She emphasized that the scale of investment is unprecedented but did not speculate on the usefulness or deployment of AI models themselves. Instead, she calculated the financial performance required for the hyperscalers to avoid bankruptcy, noting that failure to meet these targets would risk their interest payments and solvency.
This analysis places the AI infrastructure buildout in the context of historical technology booms, comparing the required growth to the US IT expansion of the mid-1990s. Wachter highlighted that achieving this level of productivity growth within a few years is challenging but feasible. The findings underscore the high stakes of the current AI investment wave and the pressure on hyperscalers to deliver significant economic returns from their capital-intensive AI data centers.
Wachter’s study projects that the hyperscalers’ AI-related expenditures will peak in 2027, with the critical productivity milestone set for 2030. The analysis was detailed in a September 15 report on technologyreview.com, providing a quantitative lens on the financial risks and potential rewards of the AI infrastructure surge.