Fed’s Barr says AI’s effect on who gets paid is still an open question

Federal Reserve Governor Michael Barr has a question about artificial intelligence that nobody can answer yet. He says it is unclear whether AI will increase returns to capital as labor’s share of income declines.

That sounds like a seminar topic. It is really a question about whose paycheck gets bigger when the robots show up: the people doing the work, or the people who own the machines.

Barr’s framing matters because the backdrop is already tilted. Labor’s share of US nonfarm business income fell to a record low of 52.8% in 2026, and AI arrives in an economy where workers’ slice of the pie has been shrinking for a long time.

What Barr is actually saying

Start with the basic split. Every dollar of income a business generates goes, broadly, to one of two places: workers through wages, or owners through profits and returns on capital.

Labor’s share measures the first bucket. For much of the post-WWII era, it sat somewhere in the 60-66% range.

Today it sits at 52.8%. That is a long slide, and it happened mostly before generative AI became a dinner-table topic.

Barr’s point is that AI could push that trend in either direction. If the technology mostly replaces workers, more of the gains flow to whoever owns the software and the servers.

If it mostly makes workers more productive, employees could capture a bigger share through higher wages. Barr’s position is that the answer remains disputed across economic and academic circles.

He has been circling this theme for months. In a speech delivered on July 14, 2026, Barr laid out how uneven the current split already is.

By his numbers, the top 20% of US households received 52% of all income. The bottom 20% received just 3%.

Two paths for AI

In that July address, Barr sketched two scenarios for how AI might play out.

The first is gradual adoption. Companies integrate AI slowly, productivity rises modestly, and workers share in the gains.

The second is faster adoption that causes significant labor-market dislocation. In this version, AI replaces workers quickly enough that the job market cannot absorb the displaced, and the gains concentrate among capital owners.

As of November 2026, Barr has acknowledged emerging signs of AI substituting for labor in specific sectors. He has also noted that broad access to these tools has not yet translated into large-scale displacement.

Why the Fed cannot fix this one

There is an awkward limit baked into all of this. The Fed’s tools are built for the business cycle, not for structural shifts in how the economy divides its output.

Fed speeches across 2025 and 2026 have emphasized that monetary policy can address cyclical problems. It cannot undo the structural distributional effects of technology.

Interest rates can cool an overheating economy or cushion a recession. They cannot decide whether a call center worker or a software vendor captures the savings from an AI chatbot.

What this means for markets and policy

There is also a demand problem hiding in the math. If the bottom 20% of households already earns just 3% of income, and AI further squeezes wage earners, consumer spending could become even more dependent on higher-income households.

For the Fed itself, the challenge is diagnostic. If AI boosts productivity, the economy could grow faster without stoking inflation. If AI instead displaces workers faster than they can be reabsorbed, policymakers could face rising unemployment that looks cyclical but is actually structural. Cutting rates would not bring those jobs back.

Barr’s own description of the evidence — some substitution in some sectors but no broad displacement — shows how murky the early signals are.

For now, Barr is doing something central bankers rarely do in public. He is admitting he does not know the answer.

Given that labor’s share already stands at a record low of 52.8%, that uncertainty is worth taking seriously.