Economists Paul Krugman, Jared Bernstein and Jason Furman say an AI bubble is possible — and how the Trump administration might respond to it.
In a Nov. 12 conversation posted to Krugman’s Substack, Bernstein said it is probably a bubble. He cited more than $1 trillion in industry capital spending since 2024, about $6 trillion planned through 2030, and revenue he put just under $300 billion, well short of the $2 trillion to $3 trillion he says is needed to justify it. Furman said he couldn’t rule it out but doesn’t trust his own analysis, recalling that he called a Tesla bubble in 2017.
Krugman said he is unsure of his own judgment.
Furman estimated about 75 percent of data center investment is imported, so Krugman noted a spending drop would be felt mostly in Taiwan and Korea. The main U.S. channel, they agreed, is the wealth effect from stocks. Furman put the boom’s boost to growth at six- to seven-tenths of a point, and Bernstein said Mark Zandi’s estimate matches at seven-tenths.
Stock busts alone rarely cause recessions, Furman pointed out, while Bernstein worried that opaque borrowing and circular financing could amplify a bust.
On the other hand, Furman said his deepest fear is that a recession becomes the moment AI-related layoffs stick, with companies not rehiring workers. “That terrifies me,” he said.
On policy, Krugman said that in what he called a contingent forecast, a bust within two years would bring no fiscal response because the White House and Congress couldn’t agree. Furman, though, said he wouldn’t predict President Donald Trump’s reaction, noting Trump acted aggressively during COVID and might want to sign stimulus checks. Bernstein said the deficit is already above 6 percent of GDP.