Fed's Goolsbee: AI data center boom 'starting to look like' demand overheating

Chicago Fed President Austan Goolsbee said Monday that the AI data center build-out is "starting to look like old fashioned demand overheating" — a diagnosis that, if it spreads within the Federal Reserve, points toward higher interest rates rather than the cuts many AI investors are counting on.

Goolsbee, who votes on the Federal Open Market Committee (the Fed's rate-setting body) starting in 2026, did not endorse a specific policy move. But his framing carries a clear implication: he is at minimum open to raising rates sooner rather than later.

His argument inverts the optimistic case for artificial intelligence. Many economists see AI as a positive supply shock — a wave of investment that raises productivity (output per hour worked across the economy) and pushes prices down. Goolsbee sees the opposite: too many resources chasing a limited number of productive uses, which is textbook demand pressure.

"In the Fed world, the bigger the hype, the more nervous I am that we are going to overheat above what the economy is capable of in the next six months," Goolsbee said.

He said he frequently hears complaints from the private sector about high costs of land, labor, utilities and raw materials that go into data centers. "That's the market telling you where the constraints are and the more we're running into constraints, the more that's demand overheating and there's no question what the Fed has to do or what any central bank has to do in a circumstance like that."

That sentence contains the sharpest signal in the speech. A Fed official saying there is "no question" what a central bank must do about demand overheating — raise rates — is not a neutral observation.

Why the supply-versus-demand distinction matters for rates

Speaking at an event hosted by the nonprofit Leadership Greater Chicago, Goolsbee laid out the conventional central banking playbook. Supply shocks — positive, like a sudden drop in energy prices, or negative, like a trade bottleneck — should be left alone to resolve on their own. Demand shocks should be addressed by changing interest rates.

The stakes are high because the U.S. economy has absorbed a run of supply shocks in recent years. Goolsbee cited ever-changing tariff policies tied to President Donald Trump's trade wars with Canada, China and other countries, and high oil prices stemming from the war with Iran.

Those shocks, he argued, should have produced isolated and short-lived price effects. They have not.

"It was supposed to have gone away by now, and we're starting to see inflation kicking up in areas that probably weren't from oil prices and tariffs, like a bunch of services," Goolsbee said. "This stuff with the AI data center build out is just, to my eyes, starting to look like old fashioned demand overheating where you get inflation coming from, [for example], everyone trying to hire an electrician at the same time and then you start seeing the prices go up."

The electrician example is not incidental. It grounds an abstract macro argument in a concrete bottleneck: skilled labor is finite, and a nationwide construction surge bids up its price.

Goolsbee added that the history of central banking argues for responding to demand-side shocks forcefully and promptly, because entrenched overheating becomes harder to fix without job losses.

"If you get the economy overheating, the conventional wisdom there is: be aggressive up front to never let the economy overheat because once it gets overheated it's extremely painful to get rid of that kind of overheating," he said. "So I think we've got to watch out in that space."

The takeaway in plain words: Goolsbee thinks the Fed should not wait for AI-driven price pressure to prove itself before acting.

When does Goolsbee actually get a vote?

Timing shapes how much weight these remarks carry. Goolsbee is not a voting member of the FOMC in 2026 — meaning that while he will take part in the discussion at the committee's final monetary policy meeting of the year next month, he will have no official say over rate moves at that gathering.

He becomes an active member next year, casting his first vote at the FOMC's January meeting. His comments now amount to a preview of the argument he will bring to the table when it counts.

Barkin: AI is a wild card pointing at employment

Another incoming FOMC voter, Richmond Fed President Tom Barkin, also spoke publicly on Monday, at a chamber of commerce event in Greer, South Carolina. He called the current combination — a healthy economy with persistently high inflation — a "relatively straightforward" case for the committee.

But Barkin framed AI as a wild card that could pull the Fed's attention toward the employment half of its dual mandate, in opposite directions.

"The path of AI feels particularly consequential. If it underdelivers, business investment could pull back, stock prices could follow, and so too could asset-fueled consumer spending," Barkin said. "The risk runs the other way, too. AI could overdeliver and displace more workers than expected, resulting in a pullback in spending. Either scenario would move attention to the maximum employment part of our mandate."

Barkin's framing matters because it treats AI not as an inflation story but as a financial-stability and labor-market story. Two incoming voters, on the same day, sketched two different channels through which AI could reshape Fed policy.

What the consumer data shows

Goolsbee's caution lands against a backdrop of stubborn inflation expectations. The Federal Reserve Bank of New York's latest monthly Survey of Consumer Expectations, released Monday, shows participants expect prices to rise 3.9% over the next year — nearly double the Fed's 2% target — while wages increase by only 2.6%.

Respondents also expressed greater concern about losing their jobs. The mean expectation of unemployment in the year ahead rose to 45.3%, the highest reading since April 2020, the depths of the pandemic disruption.

The economy itself keeps growing. Goolsbee noted that consumer spending has driven the continued expansion, with GDP rising at an annualized rate of 3.4% in the third quarter of this year. Sentiment, meanwhile, remains pessimistic — and Goolsbee argued the link between the two has broken.

"There's been a break between consumer sentiment and consumer behavior which is not just in the last three to five years, but has gotten the most extreme it's ever been in the last three to five years," he said. "It used to be that if consumer sentiment plunged, it meant in very rapid order that consumer spending was going to plunge. Consumer sentiment has been in the toilet now for a long time and consumer spending has been the main driver of growth in the economy."

He said dour sentiment has long tracked gas and grocery prices but has become less indicative of actual economic declines over time.

The catch: AI may still be disinflationary — but priced in

Goolsbee acknowledged the complicating factor in his own argument. AI could one day produce disinflation by raising productivity. Yet he said those gains may already be priced into stock prices — a development that fuels more spending by people and companies who feel wealthier.

That is the sharpest edge of his case. If AI's future productivity gains are already inflating asset values today, the build-out generates demand pressure now on the promise of supply benefits later. The Fed, in his telling, cannot wait on a payoff that markets have already cashed.

What remains unsaid is how much data would tip Goolsbee from openness to advocacy for hikes. His January vote — and whether services inflation keeps rising in areas untouched by tariffs and oil — will show whether this diagnosis hardens into policy.