Computer equipment outspent US housing by $4 billion in the second quarter

Work back through the percentages and the two lines are not mirror images. If $748 billion sits 18% below the early-2021 peak, that peak was about $912 billion, so residential investment has given up roughly $164 billion ^[1][2]. If $752 billion is 51% above where the equipment line started, it started near $498 billion and has added about $254 billion ^[2][3]. Computers gained roughly $90 billion more than housing lost ^[4]. The two moves have separate causes, too: housing has been largely frozen since 2022, when the Federal Reserve began raising rates to bring inflation down ^[11].

The crossing itself is $4 billion on a base of $748 billion, or about half a percent ^[1]. A few billion dollars of movement in either series puts residential back in front.

"We're seeing a pivotal shift in the U.S. economy: investment is shifting away from residential investment and towards computers," Shapiro said in his post, and added that "The AI investment boom is massive" ^[4][5]. His comparison uses national accounts data, and the category is described in Fortune's item as covering data centers and computer hardware ^[3]. The claim that a handful of companies are driving this comes from elsewhere in the same item: S&P Global's basket of six named firms, one of which is SpaceX ^[7]. S&P's path takes those six from $470 billion in 2025 to more than $1.3 trillion in 2027, a 2.8-fold increase in two years ^[7][5]. That 2027 figure is about 1.7 times the entire $752 billion equipment reading ^[6]. The two series count different spending.

S&P's forecast comes with a caution. The ratings firm said capex is growing faster than revenue and warned the build-out could leave overcapacity if demand does not arrive as expected ^[8]. It expects revenue to accelerate and capex to flatten in 2028 ^[9]. Until then it sees cash flow across the six as collectively negative in 2026 and 2027 ^[10]. Fortune's item uses the term operating cash flow for that measure and does not give S&P's definition.

So the crossover is real and it is thin. The concentration story is a separate question: a national equipment total tells you what was bought, not by whom, and the $1 trillion-a-year figure attributed to a handful of companies is a projection of company capex ^[12][7]. Both can be true at once, and only the equipment total is measured.