A new analysis estimates that building AI infrastructure will require about $10.3 trillion of investment through 2032. That spending equates to roughly 3.6% of GDP per year and surpasses previous major U.S. infrastructure investment waves, including railroads, highways, the electric grid and telecommunications.
Why this matters
The scale and cost of the AI buildout are too large for the biggest technology firms to fund alone, creating trillions of dollars of demand for outside capital and associated financial risks. The research presenting these findings is being shown Friday at the Brookings Papers on Economic Activity.
Financing the buildout will involve tapping bond markets, banks and private credit at very large scale, increasing overall debt demand and competing for investors’ funds at a time of rising borrowing costs.
Investment appetite versus rising rates
So far, higher borrowing costs do not appear to have significantly curbed the industry’s appetite to build. Stijn Van Nieuwerburgh, a professor at Columbia Business School and a co-author of the paper on how the AI infrastructure boom is being financed, told reporters that hyperscalers remain strongly committed to construction and that recent rate hikes and historically high long-term yields will not deter much of this development.
Morgan Stanley estimates Big Tech will need about $2.9 trillion to expand computing capacity through 2028, and the paper says more than half of that will have to come from outside investors.
Example: Meta’s Hyperion data center
Facebook parent Meta financed most of its roughly $30 billion Hyperion data center with outside investors. According to the paper, the project’s debt carries an interest rate at least 1 percentage point higher than Meta likely would have paid on its own debt, which adds an estimated more than $5 billion in costs over the life of the deal.
How risks spread through the financial system
The influx of outside capital, much of it from private credit, enables the AI buildout to scale further but spreads potential risks across the financial system via financing structures that can be opaque and difficult for investors and regulators to track.
The study estimates the new infrastructure would need to generate about $3.7 trillion in annual revenue over the next six years to deliver a 10% return — roughly 9% of GDP.
Van Nieuwerburgh warned that as the industry taps pension funds, sovereign wealth funds, private credit and other sources, the distribution of risks becomes more complex and less transparent. "These risks are coming to a pension fund near you, so to speak," he said.
What to watch next
The authors caution that a large influx of outside capital risks pushing capacity growth ahead of real demand. Van Nieuwerburgh said this could lead to oversupply — similar to cycles seen in real estate — and subsequent price declines.
Bottom line
The AI investment boom has become a major force in the broader economy: it can support growth but is placing new pressures on capital markets and the financial system. According to the paper’s authors, this buildout is by far the largest historical infrastructure expansion — about 50% larger than the second-largest, the railroad buildout.
