Spending on artificial intelligence infrastructure is on course to become the largest buildout in U.S. history relative to the size of the economy, according to a Brookings Institution study by economist Stijn Van Nieuwerburgh, CTech by Calcalist reported Sunday. The study projects $10.3 trillion of investment between 2025 and 2032 and warns that the way it is being financed could spread losses widely if AI revenue disappoints.

What Happened

The $10.3 trillion figure works out to an average of about 3.6% of U.S. gross domestic product each year over the period. By that measure the study ranks the AI buildout ahead of earlier national projects such as canals, railroads and electrification. For comparison, CTech noted that Goldman Sachs estimates U.S. AI investment will equal about 1.9% of GDP in 2026.

Spending is heavily concentrated. Microsoft, Meta, Alphabet, Oracle and Amazon are projected to invest $4.2 trillion through the end of 2029, and the study says a growing share of that is being financed with debt.

RelatedFirmus Pulls A$44 Billion ASX Listing After Weak Investor Demand

The report also tallies the effects already visible in the labor market. More than 750,000 AI-related jobs were created in the United States between 2023 and 2026, and the median advertised salary for AI-related positions on LinkedIn is about $180,000, compared with roughly $80,000 for other roles. In the Washington, D.C., area, the number of unionized electricians has risen from about 9,000 to 17,500 in recent years.

Why It Matters

The study’s warning is less about whether AI spending is too high than about how the risk is distributed. Much of the money is moving through joint ventures, private credit, securitizations and special-purpose vehicles, structures that make it hard for regulators or investors to see who is exposed. Van Nieuwerburgh frames the key question as the circumstances under which losses at individual projects could become correlated across institutions, so that one disappointing data center deal becomes a problem for many lenders at once.

The buildout is also pulling on physical resources. Data center construction is straining regional labor supplies, electricity and land, the study says, which can raise costs for other industries competing for the same electricians, power and sites and contribute to higher power prices.

The Wealth Channel

The study also points to a wealth effect. U.S. stock and mutual fund holdings totaled $63 trillion in the second quarter of 2026, nearly double their level at the end of 2022, according to the figures CTech cited. That is the period in which the AI investment cycle took off, and it ties household balance sheets to the same companies whose spending the study tracks, adding a channel beyond lenders through which a disappointment in AI returns could reach the wider economy.

What to Watch

Hyperscalers’ third-quarter earnings in October will show whether capital spending guidance keeps rising and how much of it is funded with new debt. Credit markets are the other gauge: demand for data center bonds, private credit deals and asset-backed securities tied to AI facilities will indicate whether lenders share the study’s concern about correlated losses. Power prices in data center regions and wage pressure in skilled trades will show how far the buildout’s costs are spreading into the rest of the economy.