Artificial intelligence and its footprint on the US economy
A Banca d'Italia paper quantifies artificial intelligence's macroeconomic footprint in the US economy, focusing on investment and production. The study highlights the crucial role of data centers, which attracted significant investment in 2025, boosting aggregate demand but with a limited net GDP contribution due to high import incidence.
Data centers: The AI backbone with global supply chains
The boom in IT and AI-related capital expenditure provided exceptional support to aggregate US demand in the first three quarters of 2025.
However, its net contribution to US GDP growth was more limited than headline investment figures suggest.
This is because a large share of the associated hardware is imported.
While activities like chip design, data center operation, cloud services, and model development mainly occur in the US, the fabrication and assembly of servers and accelerators are concentrated abroad, limiting domestic value creation.
Data centers are the backbone of the AI ecosystem, aggregating computational capacity for AI model training, inference, and cloud services.
They convert AI infrastructure investments into a service, generating significant demand for construction services, high-performance hardware, and ancillary inputs.
This activity is a pivotal channel through which AI investment feeds aggregate demand and contributes to GDP growth.
Revenue streams and emerging macroeconomic risks
As new AI data centers become operational and sell computational and AI services, their revenue stream will add to GDP at a scale comparable to the capital expenditure for expanding production capacity.
This is due to high utilization rates, current GPU pricing, and short pay-back periods.
However, short reinvestment cycles and uncertainty about future AI demand, while not yet a drag on growth, can fuel macroeconomic risks over the medium term.
The paper adopts a mechanical accounting approach and a short-term horizon, focusing on how the AI wave manifests in US GDP through data-center infrastructure and cloud/AI services.
The authors conclude that the financial system faces a choice between manageable adjustment costs today and potentially severe disruption tomorrow.
A complex picture, not a simple boom
This paper offers a vital, early accounting of AI's macroeconomic footprint, grounding the debate in concrete data.
The emphasis on import leakage in the AI hardware supply chain is a critical insight, tempering the narrative of a purely domestic investment boom.
Future service revenues promise further GDP contributions, but the highlighted medium-term risks from rapid reinvestment cycles demand proactive policy attention.