AI Build-Out Projected at $10.3 Trillion Through 2032

Summary

• US data-center and AI investment is projected at $10.3 trillion from 2025 through 2032.
• That averages 3.6% of GDP a year, a scale unmatched since the 19th-century railroad boom.
• The spending is absorbing labor, power and credit that other industries and homebuyers need.

The latest

US investment in data centers and AI infrastructure is projected to total $10.3 trillion between 2025 and 2032, according to estimates by economist Stijn van Nieuwerburgh published by the Brookings Institution. That averages 3.6% of gross domestic product each year. The Wall Street Journal reported that no single industry's build-out has claimed a larger share of the US economy since the railroad boom of the late 19th century.

Details

• The headline number: Van Nieuwerburgh's estimates put total US data-center and AI infrastructure investment at $10.3 trillion over eight years, or an average 3.6% of GDP annually. Goldman Sachs separately projects AI investment will reach 1.9% of GDP in 2026, a narrower measure covering a single year rather than the full build-out period.
• The crowding out: Commerce Department figures show a seasonally adjusted $37 billion spent on private data-center construction through July, about $9 billion above the same period last year. Private construction spending on everything else ran roughly $46 billion below year-earlier levels over those seven months.
• The hyperscalers: Analysts tracked by FactSet estimate capital spending at five companies — Alphabet, Amazon.com, Meta Platforms, Microsoft and Oracle — will total $4.2 trillion in the four years ending in 2029. That single-company cluster accounts for a large share of the projected national build-out.
• The debt question: A growing share of hyperscaler spending is financed by borrowing, often through off-balance-sheet entities that raise money from banks and private-credit firms with little public reporting. The Journal reported this structure carries risks for the financial sector if the AI boom turns.
• The jobs: LinkedIn estimates AI was behind more than 750,000 new US jobs from 2023 through 2026 to date. Median advertised salary for AI-related listings is around $180,000, against roughly $80,000 across all job listings. Data centers alone have added 117,000 jobs since the start of 2024.
• The price effect: Demand for data-center equipment, particularly memory chips, is creating shortages and pushing up costs for consumer technology. Prices importers paid for computers, peripherals and semiconductors were 20% higher in August than a year earlier, feeding into inflation.
• The rates link: Federal Reserve chairman Kevin Warsh named borrowing by hyperscalers as one reason long-term interest rates are higher. Those rates have made homeownership less affordable for millions of Americans, extending the build-out's cost well beyond the technology sector.
• The constraints: The scale of spending is straining the availability of labor and electricity while driving up land costs, according to the Journal. Manufacturing is among the industries being squeezed as capital, power and workers are pulled toward data-center projects.

Background

The late-19th-century railroad build-out is the only comparable episode in which one emerging industry's construction absorbed a larger share of US economic output, a comparison drawn in the Journal's reporting on the current AI spending wave.

Between the lines

Two figures sit uneasily together. Data-center construction is rising while all other private construction falls, and the borrowing behind it increasingly sits off balance sheet with limited public reporting. That combination means the economy's exposure runs through credit markets as much as through technology, which is why the Fed chairman's remarks tie hyperscaler borrowing to mortgage costs.

What's next

Watch quarterly capital-spending guidance from the five hyperscalers against the $4.2 trillion four-year estimate, monthly Commerce Department construction data for whether non-data-center building keeps falling, and import price readings for semiconductors.

Source: The Wall Street Journal (Konrad Putzier, Justin Lahart); Brookings Institution; Goldman Sachs; FactSet; Commerce Department; LinkedIn