US AI data center investment equals 3.6% of GDP over several years

The scale of this buildout exceeds anything the US economy has previously seen. Stijn Van Nieuwerburgh, a Columbia economist whose analysis was reported by heise online, calculates that AI infrastructure investment in the United States amounts to 3.6% of the country's GDP when spread over several years. That is almost double the previous record, set during the railway boom, when 2.24% of GDP was committed.

Comparison with past major cycles

To gauge the anomaly, Van Nieuwerburgh places the current trend in the context of historical network expansions. The canal, electrification, highway construction, and fiber optic booms, the last during the internet bubble, mobilized between 0.5% and 1.13% of GDP. The current buildout operates at a fundamentally different level.

The critical issue is not the amount, but the financing. Until 2024, the major players, Oracle, Amazon, Alphabet, Microsoft, and Meta, generated at least twice as much revenue as they spent. In 2026, that ratio is expected to reverse. Investments will likely exceed revenues, even though those revenues are rising sharply each year. Lacking sufficient cash flow, a growing share of funding now comes from external sources, increasing risk.

Financial models become more opaque

The economist draws parallels with earlier episodes. General-purpose technologies trigger complementary investments in vast physical networks. Returns depend on uncertain future demand, and financing needs are substantial. One notable difference is the rapid obsolescence of hardware purchased in bulk. Infrastructure financing spans long periods, while graphics cards and other components may lose value much faster. This creates a significant imbalance.

The analysis even references the subprime crisis that triggered the 2008 financial crisis, though it does not suggest imminent financial difficulties. Meanwhile, commitments are multiplying. Anthropic announced it will spend $11.6 billion over seven years on Akamai's cloud infrastructure, more than six times the $1.8 billion deal reported in May. The commitment is not unconditional. It depends on meeting specific delivery and availability requirements, and both companies can terminate it under certain conditions.

Local constraints are mounting

Resistance is forming at the local level. In Virginia, the Loudoun County board, located in the historic heart of data centers, voted to pause new siting applications. The debate centers on electricity, water, and acceptability for nearby residents. These issues are moving up the political agenda.

A general-purpose technology drives investments in a physical network whose returns depend on uncertain future demand.