How Should the Promise of Higher Productivity Growth Change the Reality of Monetary Policy Today?

May 28, 2026

At the Reykjavík Economic Conference 2026, St. Louis Fed President Alberto Musalem delivered a keynote address, “How Should the Promise of Higher Productivity Growth Change the Reality of Monetary Policy Today?” He also participated in a Q&A moderated by Claire Jones of the Financial Times. The conference in Reykjavík, Iceland, was hosted by the Central Bank of Iceland and the Center for International Macroeconomics at Northwestern University.

Full Text of Prepared Remarks

The text is as prepared for delivery.

Good afternoon. I would like to thank Governor Jónsson for the invitation to the conference and for the opportunity to address you today.I would like to thank Riccardo DiCecio, Miguel Faria-e-Castro and David Wheelock for assistance in preparing these remarks.

I will focus my remarks on productivity growth and implications for monetary policy. This topic has generated considerable interest in the United States, where excitement about artificial intelligence (AI) has reached a fever pitch. Yet productivity growth is a broader topic than just AI.

Before I get started, let me stress that these are my views and not necessarily those of my FOMC colleagues.Several of my FOMC colleagues have commented recently on these issues. For example, see Barr (2026), Cook (2026), Daly (2026), Jefferson (2025) and Waller (2025); Blanco (2026) provides a summary of the panel conversation between President Collins and President Barkin. Then-Chair Powell answered questions from the press on these topics, e.g., at the January 2026 and the March 2026 post-FOMC press conferences. Let me also note that I am an avid user of AI. I have six AI tools on the home screen of my phone. I see the tremendous impact that it can have on businesses and households. What is less clear to me is the impact that AI is having on aggregate productivity growth today and will have in the future.

Productivity growth in the United States has largely recovered from the COVID doldrums, driven in large part by firms’ investments in various automation technologies, including AI. Many observers expect productivity to continue to increase sustainably at a high rate, potentially easing inflation pressures and allowing the Fed and other central banks to lower their policy rates.

I am an AI and productivity optimist. But I live in Missouri—known in the U.S. as the “Show Me” state—where we insist on solid evidence before reaching a conclusion. To date, the data are inconclusive about aggregate productivity being in a sustained higher growth regime.

However, the demand pressures associated with the AI boom are real. We see them in the data center buildout, the demands for electricity and memory chips, and the buoyant share prices of AI companies that are helping propel consumer spending by increasing household wealth.

That brings me to monetary policy. With the real policy rate sitting below the FOMC’s notion of long-run neutral, inflation running meaningfully above target, longer-term inflation expectations drifting higher and the labor market remaining stable, I believe it would be risky to rely on the prospect of higher productivity growth in the future to solve our inflation problem today.

I’m prepared to adjust my position if the evidence becomes clear that higher productivity growth is pushing inflation lower to target. But for now, I believe a vigilant focus on returning inflation to target will best ensure success in achieving both maximum employment and price stability for the American people.

AI and Productivity

To borrow Robert Solow’s famous observation about the computer revolution in the 1980s, today we see AI everywhere but in the aggregate productivity statistics.Robert Solow. “We’d better watch out,” New York Times Book Review, July 12, 1987, p. 36. Capital investments related to AI are clearly stimulating aggregate demand, but what about aggregate supply?

As with major general-purpose technology inventions of the past, AI is expected to raise the economy’s speed limit by boosting the rate of productivity growth. In simple aggregate supply and demand terms, faster productivity growth translates into increased aggregate supply and potentially eases inflationary pressures. In theory, by lifting productivity growth, AI could help reduce inflation, allowing the Fed and other central banks to lower interest rates. At least that’s the theory.

My optimism about AI and productivity growth comes from conversations with business leaders and research into AI adoption focused on individual occupations and industries. Many companies across industries have shared with me how they are investing in AI to make their operations more efficient. St. Louis Fed employees are using AI. Some 86% of our staff have been trained to use AI and 77% use an approved AI tool in their work. I’ve seen firsthand how AI can speed up coding and other tasks. Research by St. Louis Fed economists finds that productivity has increased in occupations where AI use has become more prevalent.See Bick et al. (2025, 2026). But it’s not yet clear that aggregate U.S. productivity growth has risen to a sustainably higher rate.

Empirical Evidence on Productivity Growth

In the U.S., both labor productivity and total factor productivity, or TFP, growth have been on an upswing since 2023. AI is a plausible contributor, though unlikely the sole reason for the recent gains in productivity. For several years, firms have been investing heavily in many forms of automation technology in response to rapidly rising labor and nonlabor costs. These investments are likely a major source of increased productivity growth since the pandemic.

Labor productivity and TFP growth have both recovered to near their post-World War II averages. Since World War II, productivity growth has alternated between periods of high growth and periods of more subdued growth.

Despite a recent upswing, aggregate data do not yet offer conclusive evidence that productivity is now in a high-growth period. Instead, the data suggest the probability of the U.S. currently being in a high productivity growth period is meaningfully below 50%, representing odds much lower than a coin flip.Fernald (2014) constructed measures of TFP growth. Regular updates are available at the San Francisco Fed’s webpage Total Factor Productivity. Using the regime-switching approach of Kahn and Rich (2007), St. Louis Fed staff estimate a 25% probability that the economy was in a high-growth regime in the first quarter of 2026. Regular updates of this approach are available at Kahn’s webpage Trend Productivity Update. I believe we should base monetary policy decisions on evidence stronger than that.

It’s an open question just how soon or by how much productivity gains at the occupation, firm or industry levels show up in aggregate labor productivity or TFP growth. When a transformative technology like AI arrives, firms must make complementary investments in new software, training and processes, among other things, before the technology pays off. Resources are being spent now, but the output gains come later, so measured productivity can dip before it rises. In addition, these complementary investments and the quality improvements they enable are notoriously hard to measure.See Brynjolfsson et al. (2021) for details of how required complementary investments and measurement issues trace a productivity “J-curve” in which productivity is low or understated at first, then rises sharply and is eventually high or possibly overstated.

In the aggregate, productivity growth is an amalgamation of growth rates of both high- and low-productivity sectors. When productivity increases in one sector and the relative price of its output falls, resources typically shift to complementary sectors that may be less productive or dynamic. That reallocation of resources can limit the impact of new technology on aggregate productivity growth.See Baumol (1967) for discussion of this effect. Note that if all goods are substitutes rather than complements, the opposite effect occurs; i.e., resources flow to the most productive sector.

The combination of these effects can help explain why aggregate productivity growth may be subdued even though studies show large productivity gains at the firm or occupation level.

Productivity Growth and Inflation: What Does Economic Theory Say?

Economics provides us with useful frameworks for studying the effects of productivity growth on inflation and the implications for monetary policy. Let’s consider three situations.

A Realized Increase in TFP Growth

First, how does the persistence of realized higher productivity growth affect inflation and interest rates?

In a simple framework, an observed increase in productivity growth has two competing effects on inflation.The simple framework I have in mind is described, for example, in the textbook New Keynesian model presented in Galí (2015, Chapter 3). The deflationary effects of TFP shocks and the demand/supply decomposition of the real marginal cost response extend to more complex models, e.g., Smets and Wouters (2007). On the supply side, higher productivity directly reduces production costs. But, on the demand side, the expectation of rising income in the future can stimulate spending and investment, driving up labor demand and wages. Which of these forces dominates depends in part on the persistence of the increase in productivity growth.

If the realized increase in productivity growth is not persistent, supply-side cost relief tends to dominate, leading to lower inflation and potentially lower interest rates.For example, see Faria-e-Castro (2025), p. 33. Alternatively, by displacing labor (see Acemoglu and Restrepo, 2018), AI could lead to a decline in the labor share of income, an increase in inequality and, ultimately, put downward pressure on the natural rate of interest, r* (see Mian et al., 2025). These effects would put downward pressure on the policy rate.

However, if the realized increase in productivity is persistent, then the demand response is larger and longer-lasting, potentially to the point of overwhelming the supply-side cost relief. The result could be higher marginal costs and upward pressure on inflation and interest rates.

The prospect of persistently high future income and productivity of capital raises the natural rate of interest. Households want to borrow and firms want to invest, putting upward pressure on real rates. When productivity growth increases are persistent, avoiding higher inflation typically means the central bank must address the higher real interest rate by raising the nominal policy rate.This intuition is transparent in the textbook New Keynesian model, where the natural rate of interest is proportional to expected future consumption, income and TFP growth. If the level of TFP (in logs) follows a stationary AR(1) process, mean reversion implies that future TFP is expected to be below its current level. Hence the natural rate falls, and the central bank cuts its policy rate. If the growth rate of TFP follows an AR(1), expected productivity growth is positive and persistent. The natural rate rises and the central bank increases the policy rate.

The post-pandemic productivity recovery so far, being driven by automation and capital deepening, is less likely to be very persistent. Labor productivity growth picked up strongly in the second and third quarters of 2025 before decelerating to 1.6% in the fourth quarter and 0.8% in the first quarter of 2026.

Whether AI will be so transformative as to deliver a persistent, possibly permanent, increase in productivity growth will be a key empirical question going forward. The answer clearly has implications for the real interest rate and monetary policy.

AI as a “News Shock”

Second, aside from persistence, timing also has implications for monetary policy.

Although the effects of AI are not yet apparent in aggregate productivity data, firms and investors clearly expect large gains in profits and productivity will be forthcoming. The share prices of U.S. tech companies have risen rapidly, and investment spending on AI-related equipment and infrastructure has contributed significantly to overall real GDP growth in recent quarters. The promise of AI is thus akin to a “news shock” that has propelled demand higher today on the prospect of increased supply in the future.See Faria-e-Castro and Ozkan (2026) for details on the effects of news shocks in the St. Louis Fed DSGE model. Along these lines, Caballero (2026) constructs a model where “bubble-like” high valuations induce an income distribution and funding feedback. The interaction with a flat marginal product of capital region, due to AI capital being labor-like, can generate multiple equilibria.

Should monetary policymakers tolerate inflation resulting from upward pressure on current demand in anticipation of lower inflation when supply eventually increases? I believe that doing so could be counterproductive, resulting in higher inflation in the future as well as in the present, regardless of whether higher productivity growth materializes.

A better policy would be to lean against demand and inflation pressures today. If the anticipated productivity increase materializes and inflation pressures ease, the policy rate could then be lowered. If higher productivity growth does not materialize, inflation could remain elevated—especially if the central bank does not continue to lean against demand pressures and inflation.

Endogenous TFP: Growth vs. Stabilization

Third, it is sometimes claimed that an easier monetary policy would stimulate investment in new technologies that generate faster productivity growth and ultimately reduce inflation. Productivity improvements result from deliberate investments in R&D, innovation and adoption of new technologies. Those investments are sensitive to financing costs, which suggests that monetary policy has the potential to influence productivity growth and not merely react to it.See Bilbiie et al. (2014), Fornaro and Wolf (2022), Moran and Queralto (2018) and Ikeda and Kurozumi (2019).

However, the claim that lower interest rates today would help spur investment in new technologies such as AI and lead to lower inflation in the future requires a lot of faith: Faith that the new technology will result in higher productivity growth and faith that central bank policy positions will have a meaningful impact on investment in it.

The claim also presumes the public will continue to expect inflation to return eventually to target. In the U.S., inflation is meaningfully above target, inflation expectations have been creeping higher, and the public is highly sensitive to rising prices. In this environment, if central bankers tolerate higher inflation today based on the hope of lower inflation in the future, the people we serve may lose confidence in our commitment to see inflation return to target.

Markets are sensitive to inflation. If the public begins to question whether inflation will ever fall back to target, market participants are likely to demand higher interest rates to compensate for higher expected inflation and inflation risk. So, moving or holding policy rates too low could actually cause longer-term interest rates to rise. That would discourage investment and have detrimental effects on economic growth and employment.

History Lessons from the 1970s and 1990s

The Great Inflation of the 1970s is a stark reminder of how costly it is for a central bank to lose its inflation-fighting credibility. In the 1970s, Fed policymakers overestimated productivity growth and the economy’s growth potential. They also attributed inflation to oil shocks and other non-monetary forces. Their misreading of underlying inflation and the economy’s growth potential, the same two variables that we are debating today, allowed inflation to take hold.

Eventually, the Volcker-led FOMC accepted responsibility for inflation and took the necessary measures to bring it under control. Painful medicine was required, including a severe recession and a period of very high interest rates. Because the Fed had lost credibility on inflation, nominal interest rates stayed high long after inflation had begun to fall, resulting in sustained high real rates.

Our current environment is also often compared to that of the 1990s, when advances in information technology boosted productivity growth. At that time, the Greenspan-led FOMC recognized the increase in productivity and did not raise the policy rate even as the economy appeared to be growing above potential.In the late 1990s, the FOMC also faced the challenge of weighing the effects of increased productivity growth I described earlier: cost-reducing/deflationary versus real interest rate increasing. Then-Gov. Larry Meyer and Al Broaddus, then the Richmond Fed president, emphasized the latter. See Wolman (2026).

However, the FOMC did not ease policy either. The nominal policy rate was held above 5% even as underlying core CPI inflation declined from around 3% toward 2%, keeping the real policy rate above 3%—a level much higher than it is today. This policy posture became known as opportunistic disinflation.

At the same time, other forces were helping with disinflation, including improving fiscal balances, the Cold War peace dividend, expanding globalization in people, trade and capital flows, and a serious recession in Asia resulting in sharply lower commodity prices. Today such forces are absent or moving in the opposite direction.

The FOMC ultimately reduced the policy rate in 1998, not to accommodate higher productivity growth, but to address financial stability concerns that began with the Asian Financial Crisis and then traveled to Russia and Brazil, culminating in the collapse of the U.S. hedge fund LTCM. Less than a year after reducing the policy rate, the FOMC reversed course and started raising it in 1999 and 2000, in response to an overheating economy and rising inflation.

Conclusion

In summary, we might be in the early stages of another productivity renaissance. We see indications in micro data, but the macro data are inconclusive, and the future is highly uncertain.

Let me repeat, I am an AI and productivity optimist. But the jury is still out on aggregate productivity growth now and in the future, the real policy rate is below the FOMC’s notion of long-run neutral, inflation is running meaningfully above target, and longer-term inflation expectations have been creeping higher. Therefore, caution seems warranted in the face of upward inflation pressures from both supply and demand forces.

If the evidence becomes clear that higher productivity growth is likely to ease inflation pressures, I’m prepared to adjust my policy views. However, at present, I believe we should keep our guard up against persistent above-target inflation today, rather than base monetary policy on the hope that we will have higher productivity growth tomorrow. AI shows great promise as a transformative technology, but the risks of a miscalculation about its impact on productivity and inflation are too great.

A better approach, in my view, is to maintain a vigilant monetary policy focused on restoring price stability. By providing a stable price environment, monetary policy can best foster rising productivity growth, maximum employment and living standards for the people we serve.

Thank you.

Notes

1. I would like to thank Riccardo DiCecio, Miguel Faria-e-Castro and David Wheelock for assistance in preparing these remarks.
2. Several of my FOMC colleagues have commented recently on these issues. For example, see Barr (2026), Cook (2026), Daly (2026), Jefferson (2025) and Waller (2025); Blanco (2026) provides a summary of the panel conversation between President Collins and President Barkin. Then-Chair Powell answered questions from the press on these topics, e.g., at the January 2026 and the March 2026 post-FOMC press conferences.
3. Robert Solow. “We’d better watch out,” New York Times Book Review, July 12, 1987, p. 36.
4. See Bick et al. (2025, 2026).
5. Fernald (2014) constructed measures of TFP growth. Regular updates are available at the San Francisco Fed’s webpage Total Factor Productivity. Using the regime-switching approach of Kahn and Rich (2007), St. Louis Fed staff estimate a 25% probability that the economy was in a high-growth regime in the first quarter of 2026. Regular updates of this approach are available at Kahn’s webpage Trend Productivity Update.
6. See Brynjolfsson et al. (2021) for details of how required complementary investments and measurement issues trace a productivity “J-curve” in which productivity is low or understated at first, then rises sharply and is eventually high or possibly overstated.
7. See Baumol (1967) for discussion of this effect. Note that if all goods are substitutes rather than complements, the opposite effect occurs; i.e., resources flow to the most productive sector.
8. The simple framework I have in mind is described, for example, in the textbook New Keynesian model presented in Galí (2015, Chapter 3). The deflationary effects of TFP shocks and the demand/supply decomposition of the real marginal cost response extend to more complex models, e.g., Smets and Wouters (2007).
9. For example, see Faria-e-Castro (2025), p. 33. Alternatively, by displacing labor (see Acemoglu and Restrepo, 2018), AI could lead to a decline in the labor share of income, an increase in inequality and, ultimately, put downward pressure on the natural rate of interest, r* (see Mian et al., 2025). These effects would put downward pressure on the policy rate.
10. This intuition is transparent in the textbook New Keynesian model, where the natural rate of interest is proportional to expected future consumption, income and TFP growth. If the level of TFP (in logs) follows a stationary AR(1) process, mean reversion implies that future TFP is expected to be below its current level. Hence the natural rate falls, and the central bank cuts its policy rate. If the growth rate of TFP follows an AR(1), expected productivity growth is positive and persistent. The natural rate rises and the central bank increases the policy rate.
11. See Faria-e-Castro and Ozkan (2026) for details on the effects of news shocks in the St. Louis Fed DSGE model. Along these lines, Caballero (2026) constructs a model where “bubble-like” high valuations induce an income distribution and funding feedback. The interaction with a flat marginal product of capital region, due to AI capital being labor-like, can generate multiple equilibria.
12. See Bilbiie et al. (2014), Fornaro and Wolf (2022), Moran and Queralto (2018) and Ikeda and Kurozumi (2019).
13. In the late 1990s, the FOMC also faced the challenge of weighing the effects of increased productivity growth I described earlier: cost-reducing/deflationary versus real interest rate increasing. Then-Gov. Larry Meyer and Al Broaddus, then the Richmond Fed president, emphasized the latter. See Wolman (2026).

References

Acemoglu, Daron and Pascual Restrepo. “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment.” American Economic Review, Vol. 108, No. 6, June 2018, pp. 1488-1542.

Barr, Michael S. “What Will Artificial Intelligence Mean for the Labor Market and the Economy?” Speech delivered at the New York Association for Business Economics, New York, N.Y., on Feb. 17, 2026.

Baumol, William J. “Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis.” American Economic Review, Vol. 57, No. 3, June 1967, pp. 415-426.

Bick, Alexander, Adam Blandin, and David Deming. “The Impact of Generative AI on Work Productivity.” St. Louis Fed On the Economy, Feb. 27, 2025.

Bick, Alexander, Adam Blandin, David Deming, Nicola Fuchs-Schündeln, and Jonas Jessen. “Mind the Gap: AI Adoption in Europe and the U.S.” St. Louis Fed On the Economy, March 30, 2026.

Bilbiie, Florin O., Ippei Fujiwara, and Fabio Ghironi. “Optimal Monetary Policy with Endogenous Entry and Product Variety.” Journal of Monetary Economics, Vol. 64, May 2014, pp. 1-20.

Blanco, Amanda. “Fed Presidents: AI Not Driving Surge in U.S. Productivity – Yet.” Federal Reserve Bank of Boston, summary of the panel “A Conversation with Federal Reserve Bank Presidents” at the “2026 Technology-Enabled Disruption Conference: Shaping the Future of Finance and Payments,” on Feb. 24, 2026.

Brynjolfsson, Erik, Daniel Rock, and Chad Syverson. “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies.” American Economic Journal: Macroeconomics, Vol. 13, No. 1, January 2021, pp. 333-372.

Caballero, Ricardo J. “Speculative Growth and the AI ‘Bubble’.” NBER Working Paper No. 34722, January 2026.

Cook, Lisa D. “Opening Remarks for the ‘AI and Productivity across the Economy’ Panel.” Delivered at “The Great Realignment: Navigating AI, Demographic, and Geoeconomic Shifts,” 42nd Annual NABE Economic Policy Conference, Washington, D.C., on Feb. 24, 2026.

Daly, Mary C. “The AI Moment? Possibilities, Productivity, and Policy.” Remarks delivered at the Silicon Valley Leadership Group, San Jose, Calif., on Feb. 17, 2026.

Faria-e-Castro, Miguel. “The St. Louis Fed DSGE Model.” Federal Reserve Bank of St. Louis Working Paper No. 2024-014C, September 2025.

Faria-e-Castro, Miguel and Serdar Ozkan. “Can AI Optimism Raise Inflation? What a Standard Macro Model Says.” St. Louis Fed On the Economy, March 26, 2026.

Fernald, John. “A Quarterly, Utilization-Adjusted Series on Total Factor Productivity.” Federal Reserve Bank of San Francisco Working Paper No. 2012-19, April 2014.

Fornaro, Luca and Martin Wolf. “Monetary Policy in the Age of Automation.” Barcelona School of Economics Working Paper No. 1290, September 2022.

Galí, Jordi. Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Framework and Its Applications – Second Edition. Princeton, N.J.: Princeton University Press, 2015.

Ikeda, Daisuke and Takushi Kurozumi. “Slow Post-financial Crisis Recovery and Monetary Policy.” American Economic Journal: Macroeconomics, Vol. 11, No. 4, October 2019, pp. 82-112.

Jefferson, Philip N. “AI and the Economy.” Speech delivered at Euro 20+, Deutsche Bundesbank, Frankfurt, Germany, on Nov. 7, 2025.

Kahn, James A. and Robert W. Rich. “Tracking the New Economy: Using Growth Theory to Detect Changes in Trend Productivity.” Journal of Monetary Economics, Vol. 54, Issue 6, September 2007, pp. 1670-1701.

Mian, Atif R., Ludwig Straub, and Amir Sufi. “The Saving Glut of the Rich.” NBER Working Paper No. 26941, April 2020, revised July 2025.

Moran, Patrick and Albert Queralto. “Innovation, Productivity, and Monetary Policy.” Journal of Monetary Economics, Vol. 93, January 2018, pp. 24-41.

Smets, Frank and Rafael Wouters. “Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach.” American Economic Review, Vol. 97, No. 3, June 2007, pp. 586-606.

Waller, Christopher J. “Innovation at the Speed of AI.” Speech delivered at DC Fintech Week, Arlington, Va., on Oct. 15, 2025.

Wolman, Alexander L. “Al Broaddus, Productivity Growth and Monetary Policy in the 1990s.” Federal Reserve Bank of Richmond Economic Brief No. 26-15, May 2026.