How much has AI-related capital spending added to GDP?
Core argument: U.S. GDP was about 0.2%–0.3% larger in 2025Q2 than it would have been had business spending on data center construction, computers and peripheral equipment, and communications equipment grown in line with its 2011Q1–2022Q3 trend.
GDP was about 0.2%-0.3% larger in 2025Q2 than it would have been if businesses’ spending on data center construction, computers and peripheral equipment, and communications equipment had grown in line with the 2011Q1-2022Q3 trend. Moreover, this impact is likely understated, because existing methodologies are (probably) not fully capturing the investment being done by the five companies responsible for the bulk of the data center buildout: Amazon, Google, Meta, Microsoft, and Oracle. I estimate that U.S. GDP would be about 0.4% higher than currently reported—or about 0.6% higher than if there had been no AI boom—if the capital expenditures of the big 5 were fully incorporated into the official data. Or put yet another way, the growth in direct AI-related capex by the big 5 since mid-2023 would correspond to about 10% of the total increase in the dollar value of U.S. GDP over the past two years. Including additional capital spending on power plants and electricity generation would lead to an even larger number.

