From Servers to Rates: AI, ICT Capital, and the Natural Rate
IMF Working Papers, October 31, 2025
Source details
- Canonical URL
- From Servers to Rates: AI, ICT Capital, and the Natural Rate
Other formats
Bibliographic details
- Authors: Giovanni Melina, Stefania Villa
- Published: October 31, 2025
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9798229028851.001
Research question and approach
- Investigates macroeconomic implications of rising investment in information and communication technology (ICT)—including AI-related hardware and software—in the U.S. economy.
- Uses a structural macroeconomic model that:
- Treats ICT as a distinct type of capital.
- Explores the degree to which ICT complements or substitutes for labor.
Main findings
- Three key insights:
- Labor and ICT have historically been only moderately substitutable.
- Technological innovations that make it easier to turn ICT investment into productive capital act like demand shocks, boosting output and inflation.
- Given uncertainty about the interaction between AI-driven ICT capital and labor, different ICT investment trajectories produce materially different macroeconomic outcomes.
Scenarios and macroeconomic implications
- When ICT tends to complement labor:
- The economy experiences strong gains in output.
- Inflationary pressure increases.
- The natural interest rate increases, implying a need for tighter monetary policy.
- When ICT tends to replace labor:
- The same ICT investment path warrants a looser monetary policy stance.
Policy-relevant implications
- Monetary policy stance should account for the evolving interaction between ICT capital and labor:
- If evidence points to complementarities, policymakers should prepare for upward pressure on the natural rate and tighter policy.
- If evidence points to substitution, policymakers may need to consider a relatively looser stance.
- Recognition that technological innovations that increase the productivity of ICT act similarly to demand shocks is central for inflation and output forecasting.
Content in this bundle
- Working Paper