Analysis: the AI investment boom represents ~1% of US GDP, roughly half the GDP share during the '90s dot-com boom and comparable to the mid-2010s US shale boom
US economist Jason Furman caused a bit of a kerfuffle in 2025 when he estimated that investment in information processing equipment & software …
Context & Ripple Effects
Furman’s earlier estimate that data centers and information-processing software drove most first-half 2025 growth made AI capex central to the macro debate. This comparison places that spending at roughly 1% of GDP—material, but below the investment intensity of the dot-com era.
The size of investment and its measured growth contribution are not the same question: later bank estimates of near-zero 2025 growth contribution challenge the stronger attribution. Meanwhile, the reported productivity pickup alongside weaker entry-level hiring in AI-exposed sectors gives the debate a labor-market dimension.
First-order effects
- The analysis gives policymakers, investors and corporate planners a historical scale benchmark: AI investment is large enough to influence aggregate demand without matching the dot-com boom’s GDP share.
- It sharpens the distinction between infrastructure spending and realized economic output, especially after Furman’s earlier first-half growth attribution drew attention to data-center and software investment.
Second-order effects
- Capital-market and policy discussions are likely to focus more closely on whether AI capex is producing durable productivity gains rather than treating spending totals as evidence of economy-wide growth.
- Hyperscalers and their suppliers face greater scrutiny over the pace and financing of infrastructure build-outs as comparisons shift from headline spending to historical investment cycles.
Third-order effects
- If AI investment remains near this scale, the US economy may see a sustained infrastructure-led technology cycle whose macro payoff depends on adoption beyond the firms building capacity.
- The conflicting growth estimates suggest measurement of AI’s contribution—not just the volume of AI capex—will become a durable fault line in assessments of productivity, employment and policy.
The trend: AI is becoming a macroeconomic investment cycle, but its lasting significance will be determined by productivity diffusion rather than infrastructure spend alone.