Is government dissaving, not AI spending, widening the current account deficit?
Core argument: The U.S. current account deficit widened from about 2.2% of GDP in 2015–2019 to about 3.7% since 2023, and Julian di Giovanni attributes roughly 1.75 points of that 2.4-point gap to government dissaving, 0.5 points to lower private saving, and essentially zero to higher investment.
Gross domestic investment averaged 21.3% of GDP in the five years before the pandemic. Over 2023–2026: Q1 (the period spanning the AI capex acceleration), it averaged 21.5%, and over the past four quarters, 21.1%. The AI boom is thus not visible in the aggregate investment rate at all. To better understand why the AI boom has not had a tangible impact on aggregate investment, look at how investment has varied across sectors of the U.S. economy. The AI contribution to nonresidential investment has not been small and has been rising in recent years, but it has coincided with weakness in other sectors of the economy. The surge in factory construction that followed the 2022 industrial policy legislation has faded, and investment in structures and equipment outside the technology complex has been soft. Data centers and information-processing equipment have helped take up the slack, but not [enough] to raise aggregate investment relative to national savings [leaving the current account largely unchanged].

