AI infrastructure spending surpasses U.S. housing investment for first time

September 20, 2026 at 11:04 PM2 min

This digest was compiled by AI from multiple sources — links to the originals are below.

Inflation-adjusted spending on information processing equipment, including data centers, reached $752 billion in the second quarter, exceeding residential investment of $748 billion. The shift marks the first time AI infrastructure investment has surpassed housing investment in the U.S. economy. The milestone comes as hyperscaler capital expenditures are projected to exceed $1.3 trillion by 2027.

Key Facts

Real private residential fixed investment was $748 billion in Q2 2026, down 18% from its early 2021 peak.
Spending on information processing equipment rose 51% over the same period to $752 billion.
S&P Global projects capital expenditures from Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX will exceed $1.3 trillion in 2027.
The 30-year mortgage rate is nearly 7% as the 10-year Treasury yield hits its highest level since 2007.
Operating cash flow from the six hyperscalers will be collectively negative in 2026 and 2027, according to S&P Global.

Investment Shift

Adam Shapiro, vice president at the San Francisco Fed, called the development a pivotal shift in the U.S. economy. He noted that inflation-adjusted spending on information processing equipment now exceeds residential investment. Bureau of Economic Analysis data shows the crossover occurred in the second quarter of 2026. Residential investment has fallen 18% from its early 2021 peak, while information processing equipment spending has climbed 51%.

Interest Rate Divergence

Residential investment is more sensitive to borrowing costs, which have risen alongside Treasury yields. The benchmark 30-year mortgage rate is nearly 7%, and the 10-year bond yield has reached its highest level since 2007. AI investment has been less sensitive to interest rates, even as hyperscalers issue more debt. Google parent Alphabet reported negative cash flow earlier this year. Treasury Secretary Scott Bessent said AI companies are almost yield-agnostic in their debt issuance because they expect high returns from AI build-out.

Capex Projections

S&P Global estimated last month that capital expenditures from the six hyperscalers will exceed $1.3 trillion in 2027. The ratings firm projects capex of $870 billion in 2026 and $470 billion in 2025. S&P warned that the aggressive build-out could lead to overcapacity if future demand does not materialize. The firm sees 2028 as an inflection point, with revenue accelerating and capex flattening. Until then, operating cash flow from the six hyperscalers will be collectively negative in 2026 and 2027.