Within the third quarter of 2025, America’s largest tech companies for the primary time spent extra on capital funding than they earned from operations. The implication is that AI, a expertise with the potential to make the financial system extra productive, is, for now, absorbing assets sooner than it’s producing returns. This submit discusses how the strain between AI’s long-run promise and its short-run prices impacts the outlooks for inflation, actual exercise, and monetary stability.
Three Channels, One Framework
Drawing on my analysis, I describe three interrelated channels—inflation dynamics, structural transition, and monetary stability—by which AI bears on the financial system (see determine beneath).
Three Channels By Which Diffusion of AI Can Have an effect on the Financial system
Inflation Dynamics
Within the brief run, the diffusion of AI can reshape how rates of interest affect inflation and actual exercise. A extensively held view is that AI, by elevating productiveness, will probably be a robust disinflationary pressure. This view could in the end show appropriate, however it skips a vital step. What issues for inflation just isn’t whether or not AI raises productiveness, however whether or not it raises productiveness sooner than it will increase the prices of adopting it.
Throughout the transition, companies divert substantial assets towards reorganization, knowledge infrastructure, and integration, which may briefly elevate manufacturing prices even because the technological frontier expands. That is the so-called “productiveness J-curve,” depicted within the determine beneath.
Measured Productiveness Can Fall Throughout the Adoption Part
The results on costs, then again, are already seen in enter markets. In 2025, the foremost AI companies (Google, OpenAI, Anthropic, Meta, Amazon, Oracle) dedicated roughly $300 billion to capital funding throughout semiconductor provide chains, energy grids, and specialised labor. Aggressive funding spending continued into the primary quarter of 2026 and is projected to rise additional, including to price pressures throughout the financial system.
Latest knowledge counsel that AI-driven demand has been pushing costs up over the previous two years, and people prices are actually passing by to costs of client electronics. For instance, the costs for reminiscence chips are up considerably. A current report signifies that power consumption and costs are additionally being affected.
Structural Transition over the Lengthy Run
AI could shift the financial system’s fundamentals: the extent of potential output and the pure price of curiosity. The important query is whether or not AI generates a one-time stage shift in productive capability or a sustained acceleration in development (see determine beneath). A stage shift briefly raises the pure rate of interest throughout the transition earlier than development reverts to baseline, whereas a development acceleration raises it completely.
Quicker AI Adoption Can Sign Both That the Financial system Is Overheating or That It’s Catching As much as Its AI-Lifted Potential
Up to now, estimates of the productiveness impression span each situations, from modest positive factors of some share factors of GDP over a decade to significantly bigger results if AI augments the innovation course of itself. The vary is vast and the uncertainty is compounded by countervailing forces, together with a potential improve in market focus and shifts in family saving and spending.
Focus issues as a result of AI adoption tends to be skewed towards massive companies: if rents accrue to a handful of incumbents, the funding increase that lifts the impartial price could show narrower than combination figures counsel, and winner-take-all dynamics can also gradual the range of analysis that sustains long-run development. On the family aspect, the autumn in consumption amongst employees whose duties AI displaces could also be solely partially offset by the positive factors of these it enhances. If the latter have a tendency to avoid wasting the next share of their earnings, combination consumption could also be weaker than productiveness figures alone would counsel.
Monetary Stability
AI is rewiring the monetary system. Till lately, the foremost AI firms funded capital funding nearly completely from retained earnings, insulating the AI buildout from credit-market situations.
That modified in late 2025: capital expenditures started to exceed working money flows, and the companies raised over $100 billion of recent debt. Beneath these headline bond points lies a extra intricate layer—off-balance-sheet undertaking finance automobiles funding knowledge heart building, securitizations backed by lease money flows, and tons of of billions in ahead lease commitments that won’t seem on steadiness sheets for years. A lot of this debt relies on AI productiveness returns that haven’t but materialized. If expectations shift, the correction may journey shortly and extensively: the identical establishments—insurers, asset managers, pension funds—maintain overlapping exposures throughout company bonds, securitizations, and personal placements, so a broad repricing would hit them from a number of instructions directly.
Placing AI in Perspective
1. AI’s Centrifugal Bind
The three channels described above don’t function independently, and their interplay creates a problem in monitoring the financial system. Take into account the next situation. AI adoption initially raises prices with out elevating productiveness (the J-curve), whereas companies, shoppers, and markets already count on stable positive factors forward. Inflationary strain builds from the provision aspect: companies’ manufacturing prices rise as enter markets tighten because of, for instance, larger costs for reminiscence chips and power. Demand-side strain builds too: expectations of future productiveness maintain elevated asset valuations and help spending right this moment, earlier than the productiveness positive factors arrive. And the monetary system is uncovered: a wave of capital expenditures financed with debt relies on returns that haven’t but materialized.
The result’s what physicists would name a centrifugal bind: forces pulling outward in reverse instructions. Increased inflation may puncture asset valuations constructed on actual however distant productiveness positive factors, triggering monetary stress exactly when the supply-side payoff has but to reach. Efforts to guard monetary stability allow the fee and demand pressures to compound. Each forces are actual and hint again to the identical underlying hole between what AI is predicted ultimately to ship and what it’s truly delivering now.
2. Rethinking “Lengthy and Variable Lags”
Milton Friedman noticed that coverage works with “lengthy and variable lags”—that when a central financial institution adjusts charges, the consequences on inflation and exercise take months and even years to materialize, and the timing shifts unpredictably throughout episodes. His level was not simply that the actual results of coverage are gradual, however that they’re gradual in methods that can’t be reliably anticipated. AI unsettles this tenet in two methods.
First, AI could reshape which lags are lengthy and that are brief. Some lags could compress: sooner info flows, algorithmic pricing, and extra elastic expectations pace up transmission. Different lags could lengthen: reorganization prices and issue market frictions imply supply-side changes can take years and transfer within the incorrect route first. The outcome just isn’t a uniform shortening however a reshaping of timing and route, in contrast to something historic fashions have been constructed to deal with.
Second, even when the lags themselves don’t change, the fast diffusion of AI complicates the excellence between cyclical fluctuations and structural shifts. Throughout a serious technological transition, it turns into genuinely laborious to inform whether or not output is rising as a result of demand is overheating or as a result of the financial system’s pace restrict is increasing. The difficulty is that the info look equivalent below each interpretations, usually for years.
3. A Cautionary Story from the Dot-Com Period
The IT revolution of the 1900s provides a cautionary precedent. Within the Nineties, Fed Chairman Alan Greenspan resisted calls to tighten prematurely, betting that IT was increasing the financial system’s productive capability. He was proper. However the dot-com crash that adopted confirmed that even when the supply-side narrative is broadly appropriate, expectations can generate asset-price dynamics that create impartial monetary stability dangers. Getting the development proper didn’t defend in opposition to the bubble.
In the present day’s AI cycle options a few of the similar tensions as that episode—unsure productiveness results, issue distinguishing provide from demand, and expectations-driven asset dynamics. However it’s unfolding inside a layered and leveraged monetary system. Consequently, the trail towards an AI-driven high-productivity financial system would possibly show to be a bumpy one.
Simone Lenzu is a monetary analysis economist within the Federal Reserve Financial institution of New York’s Analysis and Statistics Group.
How you can cite this submit:
Simone Lenzu, “AI’s Macroeconomic Challenges and Guarantees,” Federal Reserve Financial institution of New York Liberty Avenue Economics, Could 20, 2026, https://doi.org/10.59576/lse.20260520
BibTeX: View |
Disclaimer
The views expressed on this submit are these of the creator(s) and don’t essentially mirror the place of the Federal Reserve Financial institution of New York or the Federal Reserve System. Any errors or omissions are the duty of the creator(s).