The Productivity Paradox: Why AI's Economic Gift May Come With an Inflation Bill

The intellectual comfort that carried markets through 2023 and into 2024 rested on a deceptively simple premise: artificial intelligence would do what technology always does—make things cheaper, faster, better. Productivity gains would anchor inflation expectations downward. The machines would be deflationary.

That assumption is increasingly fragile.

Federal Reserve officials have begun acknowledging what the data quietly suggested throughout 2024: the relationship between productivity and inflation is not linear. When productivity surges genuinely and sustainably, it raises expected future income and investment returns. Households and businesses respond rationally. They consume more today. They invest more today. Demand increases now, even as supply capacity grows. The natural rate of interest rises. This is not pathological. It is the economy working exactly as it should. It is also inflationary.

The mechanics matter because they upend three years of Fed messaging. Jerome Powell's testimony to Congress in June 2023 explicitly framed AI as a potential disinflationary force through productivity gains. Markets internalized this. Investors built portfolios around the assumption that tech-driven productivity would permit accommodative monetary policy without reigniting price pressures. By late 2024, this narrative had calcified into orthodoxy across earnings calls and policy briefings.

Then reality intruded with numbers that no longer fit the story.

Nonfarm business sector labor productivity grew at a 2.5 percent annualized rate in Q3 2024, according to the Bureau of Labor Statistics—the fastest pace since 2010. Yet inflation, measured by the core PCE deflator, remained sticky at 2.8 percent in November 2024, stubbornly above the Federal Reserve's 2 percent target. Corporate profit margins did not compress as the disinflationary narrative predicted. Instead, S&P 500 operating margins held near record levels through the third quarter of 2024. This is not the vital-sign pattern of an economy where productivity is gently pressing down on prices. This is the pattern of demand outrunning supply, with corporate pricing power intact.

The historical precedent is instructive and unkind to optimists. During the late 1990s technology boom, Federal Reserve Chair Alan Greenspan made essentially this same bet. In his July 1999 testimony before Congress, Greenspan argued that productivity gains from information technology would permit "strong growth without inflation." He was right, for a while. Inflation did remain contained through 1999 and 2000 even as growth accelerated. By 2001, the picture had darkened. Wage growth had outpaced productivity growth in late 2000. Inflationary pressures emerged across goods and services. The Fed was forced to raise rates through the cycle until the economy contracted and the dot-com recession arrived in 2001.

The parallel invites caution. We are not in the late 1990s. But we may be in a moment where AI-driven productivity is real—2.5 percent gains in business-sector productivity represent genuine improvement, not speculation—and where the demand surge it triggers is also real. The demand for energy to power AI infrastructure alone added an estimated 15 gigawatts of electricity demand into U.S. grid planning through 2026, according to Goldman Sachs research from October 2024. That is not disinflationary. That is real demand that must be met.

A National Bureau of Economic Research working paper published by researchers at the University of Chicago in September 2024 offered a more precise framework. Industries with above-trend productivity growth have indeed seen relative producer price declines. But the lag is long—effects materialize over subsequent quarters, not immediately. And the sectoral composition of the shock matters intensely. If AI productivity concentrates in energy-intensive or capital-intensive sectors, the overall price effect may be inflationary even as unit costs in those sectors fall. If AI productivity spreads broadly and permanently across the economy, demand will rise faster than cost declines can offset it.

What the Fed appears to be acknowledging in closed meetings—evident in the more cautious language from Powell since September 2024—is that policy cannot simply assume AI productivity will permit looser monetary policy. That assumption was always backwards. Productivity that raises future income and investment returns does not argue for accommodation. It argues for restraint, at least until demand and supply have realigned.

This is the inversion that policy makers never wanted to contemplate. Technology was supposed to be the escape valve for hard choices about monetary policy. Instead, it may be accelerating the very pressures the Fed spent 2022 and 2023 fighting. An economy that gets richer faster is an economy that demands more, pays more, and prices things higher—at least until the Fed can slow it down again.

The machines are not saving us from inflation. The machines are the inflation.