Can AI Optimism Raise Inflation? What a Standard Macro Model Says

The arrival of ChatGPT in late 2022 set off a wave of optimism about artificial intelligence and its potential to transform the economy. Business executives, investors and policymakers began talking about productivity gains that could rival those of the Industrial Revolution.See, for example, the predictions of Demis Hassabis, CEO of Google DeepMind, claiming that AI will have about 10 times the impact of the Industrial Revolution in a Yahoo Finance article.

Billions of dollars have since flowed into data centers, advanced computer chips, and AI infrastructure. Yet actual productivity numbers seem to reflect a more modest impact so far. Since the debut of ChatGPT, total factor productivity (TFP) growth—a measure of how efficiently an economy converts labor and capital into output—has averaged just 1.11% annually (when adjusted for utilization), below the historical average of 1.23%, according to data from the Federal Reserve Bank of San Francisco. Over the past four quarters, this slowed to just 0.32% by the fourth quarter of 2025.

This gap between AI enthusiasm and measured productivity raises an important question: Even if the productivity gains haven’t arrived yet, how does the expectation of such gains affect the macroeconomy? In this blog post, we use the St. Louis Fed dynamic stochastic general equilibrium (DSGE) model to examine this question, analyzing what economists call a “TFP news shock”—a situation in which households and firms become convinced that a productivity boom is coming, even though it hasn’t happened yet.

As our findings show, the short answer is yes: AI optimism can raise inflation, and the monetary policy response matters considerably for the extent of that increase.

The Macroeconomic Model

The St. Louis Fed DSGE model is a medium-scale DSGE model calibrated to the U.S. economy. The model features households that work, consume, save and invest. Firms hire workers and rent capital to produce output. The government collects taxes to finance its expenditures. The central bank sets the short-term nominal interest rate, which translates into changes in real interest rates because of nominal rigidities in prices and wages. For details, see the 2024 working paper by Miguel Faria-e-Castro, who used this model to analyze postpandemic inflation in a previous blog post.

How a News Shock Works

In our model, a TFP news shock captures the moment when economic agents learn that productivity will be substantially higher four quarters from now. Upon arrival of the positive news, the economy behaves as if it received a demand shock. Output growth (on a per capita basis) and inflation both rise above their steady state values (marked by the dashed horizontal line in the figures below), even though potential output hasn’t actually changed. Productivity is still the same; it’s just expectations that have shifted. (See the following figures.)

Households think of the news as a credible tip about a future pay raise. They don’t have the money yet, but they start spending anyway to smooth consumption. For firms, it is a tip about a more productive technology, for which they modestly increase investment today so as to be ready for when those productivity gains arrive.

Together, these forces produce an inflationary surge in aggregate demand—the defining feature of the news shock’s initial phase. The Federal Reserve responds to above-target inflation by raising the federal funds rate.

When the Optimism Pays Off

If the anticipated productivity gains materialize at the four-quarter mark, the economy transitions into a more benign phase. Higher TFP raises potential output, and this expansion in the economy’s productive capacity brings inflation back down toward the Federal Reserve’s 2% target. The Fed can then cut rates in response, and investment growth remains elevated as the new technology raises the returns to capital. This is the relatively happy ending—a period of demand-driven inflation followed by supply-side relief.

But the outcome depends critically on how the Fed responds during the anticipation phase. We examine a counterfactual in which the Fed fully accommodates the news shock, keeping the federal funds rate constant throughout (dotted gold line in the next set of figures) rather than raising the rate initially.

Without higher rates to restrain demand, the initial jumps in output and inflation are substantially larger. The problem compounds when the productivity gains arrive: Higher TFP raises the neutral rate, meaning that holding the policy rate at its previous level becomes increasingly accommodative precisely when the economy least needs stimulus. The result is inflation that remains persistently elevated above target even after the supply-side improvements materialize.

When the Optimism Doesn't Pay Off

The more sobering scenario is one in which the anticipated productivity gains fail to materialize; this is represented in the dashed red line in the figures below. Through the third quarter following the news, the dynamics are identical to the baseline; households and firms behave the same way because they don’t yet know the good news is false. But when the fourth quarter arrives without the expected productivity improvements, the adjustment is sharp and painful.

Output and investment growth decline abruptly as expectations are revised downward, and the wealth effect that drove the initial expansion goes into reverse. Yet inflation remains elevated because potential output never increased as expected, while the demand pressures built up during the optimistic phase persist.

The Federal Reserve, still prioritizing price stability, keeps the federal funds rate above its steady-state level even as growth stagnates. (See below.) The result is a prolonged period of weak growth and persistently elevated inflation, which is difficult for policymakers to address without accepting costs on one dimension or the other.

Conclusion

Using a standard macroeconomic model, we find that news about future AI-driven productivity gains creates immediate inflationary demand pressures that warrant a timely monetary policy response. When the anticipated gains materialize, the economy experiences stronger output growth accompanied by declining inflation, as potential output expands. But if they fail to materialize, the economy risks entering a prolonged period of weak growth and persistently elevated inflation.

One important caveat is that our model treats TFP growth as an exogenous process; that is, the growth rate of productivity is completely independent of current economic conditions. An alternative interpretation is that while a more productive technology has already been invented, firms and workers will only gradually adopt it into their workflows, with the technology’s full potential unlocked only through diffusion and adaptation. In such a framework, monetary policy may itself influence the pace of technology adoption: Tighter financial conditions that curtail investment could slow AI deployment, while accommodative policy might accelerate it. Capturing these endogenous dynamics is an important direction for future work.

Note

1. See, for example, the predictions of Demis Hassabis, CEO of Google DeepMind, claiming that AI will have about 10 times the impact of the Industrial Revolution in a Yahoo Finance article.