US Corporate Spending on AI Hardware Surpasses Residential Housing Investment

In the second quarter of 2026, inflation-adjusted corporate spending on information-processing equipment reached an annualized $752 billion, according to data from the Bureau of Economic Analysis. That figure narrowly edged out the $748 billion recorded for real private residential fixed investment during the same three-month window. The crossover marks the first time in modern United States history that businesses spent more capital building data centers and acquiring computing hardware than developers and individuals spent building homes.[1][3][4][5]

The inversion is the result of two economic curves moving in opposite directions over a five-year period. Since early 2021, residential fixed investment has declined by 18%, constrained by borrowing costs as the benchmark 30-year mortgage rate hovers near 7%. Over that exact same span, spending on information-processing equipment has surged 51%. The divergence illustrates how different sectors absorb the cost of capital. While housing construction relies heavily on debt financing and stalls when rates rise, technology infrastructure is currently expanding through a mix of massive cash reserves and debt issuance that treats high yields as a secondary concern.[1][4]

Adam Shapiro, a vice president at the Federal Reserve Bank of San Francisco, highlighted the divergence in a public analysis of the BEA data. "We're seeing a pivotal shift in the US economy: investment is shifting away from residential investment and towards computers," Shapiro wrote, adding that "the AI investment boom is massive." He noted that while the housing market remains highly sensitive to borrowing costs, the technology sector's buildout has proceeded largely unhindered by the elevated interest rate environment.[3][5]

The scale of the hardware acquisition is driven almost entirely by a concentrated group of hyperscalers. S&P Global projects that combined capital expenditures from just six companies—Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX—will climb from approximately $470 billion in 2025 to $870 billion in 2026. By 2027, that combined outlay is expected to exceed $1.3 trillion. The spending flows directly into the physical stack required to train and run large language models: specialized graphics processing units, high-density server racks, cooling systems, and the vast concrete shells required to house them.[1][3]

Because the anticipated returns on artificial intelligence infrastructure are perceived as generational, the companies funding the expansion are willing to borrow at current rates to maintain their pace. US Treasury Secretary Scott Bessent recently described corporate debt issuance for AI projects as "almost yield-agnostic." Firms are prioritizing capacity over capital efficiency, calculating that the cost of missing the technological transition far outweighs the penalty of 2026 borrowing rates.[1][5]

The sheer velocity of the capital reallocation is altering macroeconomic growth models. Torsten Slok, chief economist at Apollo Global Management, calculates that data-center capital expenditure will rise from 1.4% of US gross domestic product in 2025 to 3.1% by 2027—an acceleration of roughly 0.85 percentage points per year. "The AI cycle is building at close to twice the pace of the housing boom at its fastest," Slok noted. That rapid expansion means that information-processing investment is no longer just a technology sector metric; it has become a primary engine of national economic growth.[3]

Yet the physical footprint of the buildout is beginning to generate friction at the local level. While a housing development visibly adds to local inventory, a 100-megawatt data center draws heavily on regional power and water grids without offering the same immediate civic utility. Recent polling indicates that 64% of registered voters would be less likely to support a political candidate who favors a new data-center project in their community.[1][4]

The durability of the crossover depends on whether the technology sector can generate enough revenue from AI services to justify the ongoing infrastructure costs. If the Federal Reserve lowers rates and the housing market thaws, residential investment could reclaim its historical position as the dominant capital sink. Until then, the marginal dollar in the US economy is flowing toward servers and cooling systems rather than timber and drywall.