Financing the AI buildout projects that AI investment in data center buildings, power systems, networking infrastructure, and specialized chips and other equipment will total an enormous $10.3 trillion from 2025 to 2032, or an average of 3.63 percent of U.S. gross domestic product per year, according to Stijn Van Nieuwerburgh, Columbia University professor and author of Financing the AI Buildout, a study published by Brookings.
“The projected buildout would be larger relative to the economy than the major U.S. canal, railroad, electrification, highway, and telecommunications investment booms,” says Van Nieuwerburgh.
source: Brookings, Financing the AI Buildout
And, as many note, opacity is growing as hyperscalers increasingly shift debt financing from their own balance sheets to third parties.
Morgan Stanley estimates that more than half of the roughly $2.9 trillion required to meet hyperscalers’
Incremental compute needs over 2025–2028 will come from outside capital, the study says.
Hyperscalers often direct a large share of their internal capital toward IT equipment, while data center shells, power infrastructure, and related real estate are financed through project-level debt, leases, and other asset-backed structures, the study says.
The upshot is that leverage is removed from hyperscaler balance sheets and shifted elsewhere. The central issue is not simply that AI infrastructure is exposed to technological and operating risks, but that the sector’s financing structure can transmit and amplify those risks, the study rightly notes.
The severity of any adverse shock will therefore depend not only on the underlying economics of AI demand, but also on where leverage resides and how losses are allocated across tenants, asset owners, and creditors, the author says.
“The relevant question is therefore not whether AI infrastructure is already systemically risky, but under what conditions project-level losses could become correlated and propagate across firms and financial institutions,” Van Nieuwerburgh argues.
All that noted, looking at Price/Earnings-to-Growth (PEG) ratios might suggest that some parts of the AI value chain might be considered undervalued, relative to their growth rates. Alphabet, Amazon and Nvidia provide cases in point.
A PEG ratio of 1.0 implies that a particular equity is valued at market averages. Conversely, PEG ratios above 1.0 imply high valuation relative to the overall market. PEG ratios below 1.0 suggest a particular firm is undervalued, relative to its growth rate and valuations of all other public firms in the market.
