Saving for the Machines: How AI and Automation Will Reshape Capital, Growth, and Wealth

 As artificial intelligence and advanced automation rapidly evolve, a defining economic question of our era is taking shape: Who will save for the machines?

The upcoming wave of economic transformation—fueled by software, massive data centers, power plants, and specialized robotics—requires unprecedented levels of capital investment. According to recent projections, AI campuses alone are expected to absorb a staggering 3.6% of GDP per year between 2025 and 2032. Because much of this capital depreciates rapidly (with high-end GPUs lasting only three to six years), funding this technological revolution demands a massive mobilization of savings.

In a comprehensive research paper titled Saving for the Machines, Jesús Fernández-Villaverde (University of Pennsylvania, CEPR, NBER) explores the macroeconomic mechanics of this transition. Utilizing a neoclassical growth model integrated with the task framework of Acemoglu and Restrepo, the paper investigates what happens as the economy races toward full and partial automation by 2045.

1. The Dual Engine: Task Structure and Capital Supply

To understand how the economy adapts to automation, Fernández-Villaverde highlights two critical margins:

1. Task Structure: Determines which tasks machines can execute, how efficiently they perform them, and how easily labor and machines can substitute for one another. This dictates the return on accumulating machines.

2. Capital Supply: Captures how households adjust their savings in response to changing economic conditions.

The paper contrasts optimal household saving (where capital supply is perfectly elastic in the long run) against a fixed saving-rate benchmark (such as keeping the savings rate locked at the historical U.S. average of 23%).

2. Key Findings: The Great Growth Boom and the Capital Share

The quantitative calibration (anchored to U.S. data for 2025 with a transition window to 2045) reveals dramatic shifts in economic trajectories depending on the elasticity of substitution ($\sigma$) between tasks:

• The Power of Optimal Saving: With an elasticity of substitution of $\sigma = 0.2$ (signifying strong complementarity between tasks), long-run growth reaches 11.7% under optimal saving, compared to a mere 2.2% if the saving rate remains locked at 23%. When $\sigma = 1$, growth skyrockets to 21.9%.

• A Capital-Share Driven Surge: Rather than households becoming drastically "thrifter," the dramatic rise in savings—surging from 23% to over 54%–57% of output—is driven by a massive expansion in the capital share. As automation takes over, the capital share of income climbs from roughly one-third toward one. Because capital captures a much larger slice of a vastly expanded economic pie, investment naturally explodes.

• The Danger of Bottlenecks: Strong complementarity ($\sigma = 0.2$) paired with a rigid, fixed saving rate can choke off the growth boom, causing the economy to drift back toward baseline productivity growth. The binding constraint is often not technology, but the supply of capital.

3. Wages, Inequality, and "Ricardo’s Rent"

The transition to an automated economy carries profound distributional consequences:

• The Initial Wage Dip and Recovery: Under full automation, wages initially face downward pressure as machines absorb human tasks. However, the long-run outcome depends heavily on task elasticity. With low substitutability ($\sigma = 0.2$), tasks that only labor can perform act like scarce fixed factors, generating large "Ricardian rents". Consequently, wages eventually recover and can rise well above historical trends.

• The Hand-to-Mouth Divide: While capital owners reap extraordinary rewards (with certainty equivalent welfare gains often exceeding 100% to 200%), hand-to-mouth households who rely entirely on labor income face severe vulnerability. Simulations incorporating hand-to-mouth consumers indicate they can lose 40% to 65% of their consumption relative to trend during the transition.

• Targeted Compensation: Model simulations show that keeping labor income strictly on trend via policy interventions requires temporary transfers that peak at around 4% of GDP ($\sigma = 1$) to 20% ($\sigma = 0.2$) around the year 2040, after which economic growth outpaces wage declines and the need for compensation fades.

4. Financial Markets and Asset Prices

The macroeconomic shock of automation also fundamentally alters financial markets:

• Exploding Long-Term Yields: Because markets anticipate an aggressive growth boom, the 10-year yield jumps sharply from around 5% to between 9% and 14% (and even higher for longer maturities).

• Human Wealth vs. Equity: Human wealth drops precipitously by roughly two-thirds as labor's share of income shifts. Conversely, the stock market experiences a massive quantity boom driven by intense capital accumulation, with equity shares of wealth surging past 95% by 2045.

Conclusion

Jesús Fernández-Villaverde’s Saving for the Machines offers a rigorous quantitative roadmap for an AI-driven future. It demonstrates that funding the infrastructure of tomorrow—from power-hungry data centers to advanced robotics—depends entirely on how households and financial systems mobilize savings. While the technological potential for unprecedented economic expansion is immense, managing the distributional divides and capital supply bottlenecks during the transition will define the ultimate success of the machine age.