U.S. spending on information processing equipment reached $752 billion in the second quarter, edging past real private residential fixed investment of $748 billion, according to federal data reported by Fortune on Sept. 20. The shift marks the first time computing investment has surpassed housing investment, with residential investment down 18% from its early 2021 peak while computing spending has risen 51%. The milestone arrives as opposition to data centers intensifies. In Utah, residents are challenging a county’s approval of a $10 billion data center campus, and El Paso adopted new rules for data centers this summer.
Investor Daniel Kaufman, founder of the permanent capital holding company Kaufman & Company, said Wednesday that this spending imbalance should change how AI campuses are built, with housing for the people who run them planned from the start instead of left to chance. “For the first time, this country is putting more money into the machines that run AI than into the homes people live in,” Kaufman said. “I’m not here to fight that capital. We develop data centers ourselves. But every campus needs electricians, operators and their families living within a reasonable drive, and right now almost nobody is budgeting for those homes. That is how you lose a town hall, and it’s how you lose a workforce.”
Kaufman & Company has a stake on both sides of the trade. ServerDomes Global, a Kaufman Ventures portfolio company, announced plans in July for geodesic dome data centers in Utah and El Paso, Texas, including an El Paso site spanning 24,000 acres. The company has said the domes are designed to use about 90% less water than conventional facilities. On the housing side, the platform includes DK Homes, Design Build Affordable and the factory-built Mr Good Container Homes, and works with Oldivai, an aligned, independently led workforce housing partner.
“Communities are right to ask what they get in return for hosting this infrastructure,” Kaufman said. “Our answer is that we don’t want a blighted site with a fence around it. We want to build housing and other uses around the domes, and we want the people who keep those buildings running to be able to live near them. An investor who treats the housing as somebody else’s problem is going to find the power, the permits and the people all harder to get.”
The implications extend beyond one company. As AI infrastructure spending continues to outpace residential investment, the ability to site and staff data centers may hinge on whether developers address the housing needs of the workforce. Community resistance, like that seen in Utah and El Paso, often stems from concerns about strain on local resources, including housing. By integrating workforce housing into campus plans, developers could mitigate opposition, secure necessary permits, and ensure a stable labor pool. Kaufman’s approach suggests that the next wave of AI campuses may need to be mixed-use developments, blending computing facilities with residential components. This strategy could become a competitive advantage as localities demand more benefits from hosting large-scale infrastructure. The shift also highlights a broader challenge: the U.S. is investing heavily in the machines of AI but underinvesting in the homes of the people who operate them. Balancing both may be essential for sustainable growth in the sector.
