Kevin Warsh is right about Fed reform - but his inflation solution is a trap

MW Kevin Warsh is right about Fed reform - but his inflation solution is a trap

By Jai Kedia

Warsh's belief that AI is a guaranteed disinflationary force could trigger premature rate cuts

Kevin Warsh is President Donald Trump's nominee to chair the Federal Reserve.

Does a 2% inflation target make sense in a higher-productivity economy?

Kevin Warsh has built much of his case for cutting interest rates on a predicted artificial-intelligence productivity boom. In a Wall Street Journal op-ed, the nominee for Federal Reserve chair wrote that AI "will be a significant disinflationary force." Moreover, Warsh claims that these AI productivity gains will give the Fed leeway to cut rates.

While Warsh's views received some attention at his nomination hearing before a U.S. Senate committee last month, they have not been sufficiently scrutinized as a basis for monetary policy.

There is a great deal worth supporting in Warsh's reform agenda. He is right that the Fed's balance sheet, expanded for crises long past, should be reduced. He is right that forward guidance has corroded the central bank's discipline rather than sharpened it. He is right that the Fed has strayed well beyond its purview and should revert its focus to price stability and full employment. I have argued versions of each of these points elsewhere. The AI productivity argument is different. It points the wrong way.

What Warsh is offering is a forecast dressed as a framework. He claims that AI will raise productivity, that such productivity gains will be disinflationary, and that the Fed should cut rates accordingly. The framework question - how the Fed should respond if productivity rises - is one thing. The forecast question - will it, when, and by how much - is something else. Warsh acknowledges little uncertainty on the second question. Professional estimates acknowledge a great deal.

The range of credible estimates on AI's productivity impact span roughly an order of magnitude. Economist Daron Acemoglu puts the total factor productivity gain at no more than 0.66% over a decade, or about 7 basis points per year. In their research, economists Philippe Aghion and Simon Bunel, drawing on prior general-purpose technology waves, estimate somewhere between 0.8 and 1.3 percentage points annually. Goldman Sachs has projected 1.5 percentage points annually, while AI insiders themselves project considerably more.

The net inflation effect is even less settled. Fed Vice Chair Philip Jefferson noted in a November speech that AI's near-term effect on prices is not unambiguously downward - the technology is currently a positive demand shock, with capital expenditure on data centers, electricity infrastructure and equipment running well above trend, before it becomes a positive supply shock to consumer goods and services. Even granting Warsh's productivity story, the timing matters.

To be clear, there is a version of the productivity-and-prices argument that holds up. If AI represents a permanent positive supply shock, the right response is not to cut rates today on a forecast but to revisit whether a 2% inflation target makes sense at all in a higher-productivity economy.

A monetary framework that lets falling production costs show up as lower prices for consumers, rather than offsetting them with compensating stimulus, has intellectual support and is worth discussing. But it is a long-run framework reform. It is not a case for cutting the federal-funds rate target in the near future. The version that justifies near-term cuts, presumably the one Warsh is making, is the weaker one.

From the archives (February 2026): Kevin Warsh needs to keep the Fed out of politics - no matter what Trump says

The problem is not the substance of Warsh's AI view. It is the method.

Even if Warsh holds these views as chair, his individual judgment is unlikely to drive overall rate decisions much. Stephen Miran's appointment to the Federal Open Market Committee last September was widely seen as a way to operationalize the president's call for sharp cuts. Indeed, Miran has been the lone dissenter (always dovish) at every meeting since his appointment. But these dissents have never mirrored Trump's desire for a massive, 200- to 300-basis-point decline in the Fed's target rate. And the other 11 members never followed Miran. Instead, Miran is now softening his stance. Appointees adapt to the committee's norms, and the committee's norms dampen the influence of any one voice.

The problem is not the substance of Warsh's AI view. It is the method. Anchoring a policy stance on a single-point estimate from a wide distribution of credible forecasts is an exercise in discretion. The AI-productivity thesis stands out as the one piece of Warsh's broader agenda that points discretionary. The rest - balance-sheet discipline, less forward guidance, a narrower mandate - pulls in the other direction. That asymmetry will invite scrutiny on its own, especially under the broader political context of the president's desire for sharp, immediate rate cuts.

If productivity gains do not arrive on the timeline the argument assumes, the cost is not just to the argument but to the Fed's credibility. The 2021 "transitory" episode is a relevant cautionary tale. The FOMC set the target rate wrong because it adopted a forecast - that pandemic price pressures would prove temporary - and built policy around it.

Ordinary Americans are still dealing with the fallout from that policy choice, and the Fed's reputation has suffered serious damage. Its credibility will be harmed further if the chair offers policy prescriptions based on a subjective assessment of economic conditions that don't materialize.

Fortunately, there's a simpler way to deal with evolving economic conditions, especially under uncertainty. A rules-based framework accommodates a productivity boom automatically.

Under such a system, the FOMC publishes an arithmetic formula linking the interest-rate target to key macroeconomic variables such as inflation and employment. If AI delivers, the reaction function adjusts and rates come down without anyone having to correctly forecast OpenAI's or Anthropic's product roadmap. If it underdelivers, rates hold. The committee does not have to bet. This is the reform Warsh's broader agenda points toward - but that his AI argument cuts against.

Discretion is the enemy here - not Warsh, and not the AI thesis. The Fed should commit to a policy rule and let the data on AI productivity speak for itself.

More: AI is 'absolutely useless' at forecasting inflation. This proven model is 12 times more accurate.

Also read: Kevin Warsh will be the richest Fed chair ever. Just how rich - he isn't saying.

-Jai Kedia

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

May 11, 2026 14:47 ET (18:47 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.