Artificial Intelligence, Structural Shifts, and Monetary Policy
Remarks by Lael Brainard
Distinguished Fellow, Georgetown University Psaros Center
Senior Fellow, Harvard Kennedy School Mossavar-Rahmani Center
at the College of Central Bankers (CCB) and Bloomberg Symposium
Paradigm Shift in the Global Economy
8:45 a.m.
December 3, 2025
Economic Transformation in the AI Era
The U.S. economy is at the threshold of a profound structural transformation. Artificial Intelligence (AI) has the hallmarks of a general-purpose technology capable of lifting productivity growth for a decade or more. While diffusion is in early stages, the promise of AI is already altering capital allocation and the relationship between growth and employment.
Conceptualizing the AI boom in terms of capital-based unbalanced growth can help to explain recent jobless growth and the crowding out of investment in non-AI sectors. In a capital-based unbalanced growth framework, the elevated demand for capital in the leading AI sector raises financing costs and crowds out investment in other sectors, especially capital-intensive non-AI sectors such as manufacturing. This AI investment drives strong growth that is jobless in the early stage because AI frontier firms do not scale up their workforces proportionately, and other firms postpone hiring due to AI uncertainty.
A Perfect Storm of Supply Shocks
Following several decades when demand shocks prevailed, a perfect storm of supply shocks has dominated the last five years. First, COVID fractured global supply chains. Then Russia’s invasion of Ukraine triggered a commodity shock that compounded the disruptions.
No sooner had these disruptions dissipated than the administration implemented a nearly ninefold increase in average tariffs, along with immigration policies that slowed labor-supply growth. The resulting stagflationary shock meant that firms that had planned around a continued decline in inflation instead confronted renewed cost pressures and supply-chain uncertainty, while households saw affordability deteriorate.
Supply shocks of this nature pose special challenges for monetary policy, which is better-suited for demand stabilization. Tariffs push inflation and unemployment away from their desired levels in opposite directions, making it difficult for the federal funds rate to improve one without exacerbating the other. Instead, the executive branch is best positioned to address the inflationary pressures associated with the tariffs without weakening labor markets. The tariffs were imposed through executive authority, which makes them straightforward to unwind.
AI as a General-Purpose Technology
Layered beneath these disruptions, a transformative supply-side shift is underway in the U.S.: AI is altering the deployment of capital and labor across and within sectors.
The large language models (LLMs) associated with major productivity gains are exceptionally capital-intensive. They require unprecedented investment in compute power and data-center infrastructure. In the current race to achieve human-scale AI, annual AI-related investment could average up to $1 trillion per year over the next five years by some estimates.[1] This is driven in part by depreciation cycles for cutting-edge chips in frontier training models that are as short as two years. This could diminish if technological progress reduces compute intensity or increases efficiency, business models shift away from LLM-centric architectures, or demand for AI does not meet expectations.
Illustrative Framework: Capital-Based Cost Disease
An adaptation of the Baumol cost-disease framework can help illuminate the AI race, where unbalanced growth features rising capital costs, instead of rising labor costs, as AI opportunities crowd out investment opportunities in the rest of the economy.[2] Here’s how it works: As capital markets bet big on AI-related investments, non-AI manufacturing and other sectors face tighter financing conditions as they compete for capital.
Whereas in the classic Baumol model, the lagging sector experiences wage pressures due to the rising labor productivity in the leading sector, today’s early-stage AI boom creates a capital-based cost disease, where slower-productivity sectors are squeezed by rising costs of capital. The most capital-intensive industries outside AI—such as non-AI manufacturing—may feel this pressure most acutely, at a time when these industries are already contending with higher cost inputs due to tariffs.
Understanding Today’s Jobless Growth
Extending this framework to allow for substitution between labor and capital helps illuminate how the AI boom could be driving recent jobless growth. AI-related investment powered one fifth of the robust 2.9 percent annualized growth in private domestic final purchases in the second quarter, at a time when employment growth fell to half its earlier pace.
While tariff uncertainty has been the focus of most commentary on the soft job market, increasingly it appears AI is also dampening hiring. Starting in July 2025, each Federal Reserve Beige Book has flagged the connection between AI and subdued hiring. Here’s how the November 2025 Beige Book described it: “A few firms noted that artificial intelligence replaced entry-level positions or made existing workers productive enough to curb new hiring.”[3]
Firms at the AI frontier are scaling rapidly without proportionate workforce expansion, while firms in lagging sectors—such as manufacturing—appear to be pausing hiring as they evaluate how AI adoption will change their capital and workforce needs. Preliminary research suggests that AI uncertainty is affecting sectoral employment in a way that is correlated sectoral exposure to AI.[4]
This high-investment-low-hiring growth is consistent with a capital-based unbalanced growth model where capital substitutes for labor in the higher productivity AI sector, or where there is a continuum of tasks, and capital investments in the high productivity tasks lead to labor substituting productivity growth.[5]
Policy Implications
The primary policy levers to address the distributional implications of capital-based unbalanced growth lie with the executive branch: tax policy, workforce development, and targeted investment support for critical sectors that AI is not lifting.
These developments could also have implications for financial stability oversight. The AI investment boom is exhibiting an all-too-familiar mix of stretched valuations, hidden leverage, interconnectedness, and liquidity mismatches. These investments face considerable uncertainty surrounding future monetizable demand or emergence of low-cost business models from China or elsewhere. A core group of AI-oriented firms have driven much of this year’s equity market gains, making the market more vulnerable to a reversal. Large amounts of debt—some of it in off-balance-sheet SPVs or embedded in private credit—are tied to the AI cloud buildouts. Some of this relatively illiquid debt is likely included in redeemable-on-demand investment funds. And the circular interconnections between frontier model developers, data-center operators, advanced chip designers, and lenders create channels through which stress could propagate.
For monetary policy, the implications of the AI boom are ambiguous. AI diffusion will reshape the supply side over years, not months. While the elevated investment demand associated with the AI race may be putting upward pressure on the neutral rate (along with higher fiscal deficits), that directional information does not pin down the appropriate level of the policy rate today, at a time when inflation has been running above target for four years, and unemployment is moving higher. Nonetheless, central bankers must understand these shifts in the real side of the economy to form accurate judgments about future productivity growth, the output gap, and the dynamics of inflation and labor turnover.
Near-Term Policy Outlook
In the near term, the Federal Reserve has one instrument for two very different economies. While tighter policy could help cool froth conditions at the AI frontier, it could risk tipping the labor-market into a downward spiral in the rest of the economy. This challenge echoes, but does not replicate, the dilemmas of the late-1990s dot-com era. Today’s conundrum is magnified by stale data, lagged tariff price pressures, and uncertainty about future policy—from potential litigation-driven changes to tariffs to a repeat of fiscal transfers to households.
These crosscurrents have contributed to unusually high dissension within the FOMC. For the final meeting of the year, policymakers must rely on stale data that extend only through September on inflation and the labor market. Members who are inclined to hold rates steady can point to September’s rebound in payrolls and labor force participation and steady initial claims, along with the lack of progress on high inflation. Those favoring a cut can emphasize the four consecutive months of gradually rising unemployment as evidence that labor demand has softened more than labor supply.
The most recent comments by leadership clearly hold open the door for a cut. Recent commentary indicates there are sufficient voters who are open to a cut that a majority can be achieved. But the presence of a few voters with strongly held views on both sides suggest dissent is likely.
Conclusion
Across these developments, one fact is clear: the diffusion of AI is driving structural changes that are only beginning to unfold. In this early phase, conceptualizing the AI boom as capital-based unbalanced growth can help to explain recent jobless growth and the crowding out of investment in non-AI sectors. For all of us, the central task is assessing how AI-driven capital deepening, workforce shifts, and widening sectoral divergence are reshaping the economy’s trajectory in real time.
[1] Bloomberg, Victor Swezey, November 10, 2025. “AI’s $5 Trillion Cost Needs Every Debt Market, JPMorgan Says.”
[2] Baumol, William J. 1967. “Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis.” American Economic Review 57(3): 415-426.
[3] Federal Reserve, Beige Book, November 26, 2025. https://www.federalreserve.gov/monetarypolicy/beigebook202511-summary.htm
[4] https://www.escoe.ac.uk/events/the-macroeconomic-effects-of-ai-uncertainty/
[5] Karabarbounis, Loukas, and Brent Neiman. 2014. “The Global Decline of the Labor Share.” Quarterly Journal of Economics 129(1): 61-103; Oberfield, Ezra and Devesh Raval. 2021. “Micro Data and Macro Technology,” Econometrica, 89(2): 703-732; and Acemoglu, Daron, and Pascual Restrepo. 2018. “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment.” American Economic Review 108(6): 1488-1542.