K en Griffin wasn’t buying the AI panic. At Davos in January, the billionaire founder of Citadel, the Miami hedge fund giant with $68 billion in investment capital, dismissed artificial intelligence’s output as “garbage.”

Then this month, Griffin did a 180 . He watched AI agents do complex work in hours that once took Citadel employees weeks or even months. Citadel’s entire business is built around hiring brainiacs. More than 40% of its employees hold advanced degrees, including about 270 Ph.D.s across 40 fields. These are some of the highest-paid workers in America –the median annual compensation for software engineers at Citadel is more than $500,000 – and software that can replace even part of that labor could save firms like Citadel enormous amounts of money. Griffin still said he went home depressed because machines were starting to do work that once only those people could do.

Economists may soon face a strange problem. Businesses grow. GDP rises. Profits stay strong. But the jobs don’t come along for the ride. If AI allows companies to produce more with fewer workers, America could end up looking richer on paper while millions of households feel poorer in real life. An economy with rising GDP and 8% unemployment would have sounded implausible a few years ago. With each passing day, it sounds a little less so. If that’s where the economy is heading, economists may have to rethink whether growth alone still tells us the economy is healthy.

Since the Great Depression, GDP has been the main measure of economic health. Economist Simon Kuznets , who would receive a Nobel Prize for his work in 1971, developed the metric in the 1930s while working with the U.S. government to track the collapse. When GDP rises, the economy is considered to be growing. When it shrinks for long enough, the thinking goes, the economy is probably in a recession or close to one. It’s not quite that black and white because the official call is made by the National Bureau of Economic Research and includes other factors, but the basic framework has remained intact for decades. Growth and recession aren’t supposed to happen at the same time.

Throughout modern American history, recessions have arrived with brutal regularity. From 1950 through 2010, the U.S. endured 10 recessions , or roughly one every six years. The economy contracted in 1953, 1958, 1960, 1969, twice during the inflation and oil shock of the 1970s, again in the early 1980s when the Paul Volcker Federal Reserve crushed inflation with punishing interest rates, then during the savings-and-loan crisis, the dot-com bust and finally the housing collapse in 2008. The details changed, but the broad pattern stayed the same. Corporate profits fell and with them so did GDP. Americans lost jobs and businesses failed. The economy looked sick because the economy was sick.

Then something changed. Outside of the brief Covid collapse, the United States hasn’t experienced a traditional recession since 2008. The longest expansion in modern U.S. history stretched from June 2009 until the pandemic shutdowns 11 years later. Since then, the economy has repeatedly bucked recession models. Massive government stimulus, years of near-zero interest rates, globalization and the growing dominance of technology firms helped keep growth alive. But even as GDP and stock prices climbed, wealth inequality widened as housing, healthcare and education costs rose faster than most paychecks. The old signals stopped lining up the way they once did. AI could widen that disconnect even further by allowing companies to grow without needing nearly as many workers.

Technology has always destroyed some jobs. Farm equipment reduced the need for manual labor. ATMs reduced the number of bank tellers. Telephone operators disappeared. But there was usually somewhere else for workers to go. New industries appeared. New jobs came with them.

AI could be different because it is moving into so many kinds of work at once. It’s already writing code, reviewing contracts, handling customer service and analyzing spreadsheets. Many of those jobs were long considered difficult to automate.

Companies are already starting to test what an AI-heavy workplace might look like. Meta is cutting 8,000 positions while Mark Zuckerberg pours billions into AI. Block, the parent company of Square and Cash App, eliminated more than 4,000 jobs after Jack Dorsey said the technology had changed what the company needed from humans. Standard Chartered, the British bank, expects AI and automation to help cut more than 7,000 “lower-value human capital” roles by 2030.

Not every “AI layoff” is really about AI. Companies overhired, investors want lower expenses, and executives now have a convenient scapegoat for job cuts. Still, more companies are starting to realize they may need fewer workers than they once thought.

Michael Madowitz , principal economist at the Roosevelt Institute, a Washington think tank focused on economic policy, says economists don’t have enough catchphrases for every strange state the economy can reach. The term “stagflation” only became common after the 1970s proved inflation and unemployment could rise together. AI could create its own mismatch of strong growth and high unemployment at the same time.

Madowitz isn’t predicting a jobless future. He’s saying its time to throw out your Econ 101 textbooks because the old way of judging the economy may stop making sense. Roughly two-thirds of national income has historically gone to workers through paychecks, with owners taking much of the rest through profits. Many economic models simply assume that split because it has stayed fairly stable for so long. But that balance comes from history, not a law of nature. If AI allows companies to produce more with fewer workers, a larger share of the gains could flow to owners and a smaller share to employees.

A strong economy with a weak labor market would be hard to ignore. Unemployment above roughly 5% already makes economists queasy. Add strong GDP growth driven mostly by profits and rising inequality, and the picture changes. That’s not to mention the greater societal implications. “You could be looking at healthy GDP growth here,” Madowitz says, “but this is not a healthy economy.”

There is a more reassuring version of this story.

Joe Brusuelas , chief economist at RSM US, an audit, tax and consulting firm focused on middle-market businesses, remembers his mother working as a telephone operator, connecting calls by hand. Advances in telecommunications eventually made that job obsolete. She moved into health care. The economy kept growing.

Brusuelas believes that pattern still holds. Every major technological leap has created fears that workers would become obsolete, and so far new industries and jobs have continued to appear. He’s skeptical AI will force a complete rethink of recessions anytime soon.

He also warns against swallowing every AI claim whole. The technology is real, but so is the hype. “Don’t believe the hype,” he urges.

AI hasn’t yet delivered an economy of booming GDP and soaring unemployment. Some layoffs blamed on AI probably have more to do with overhiring or pressure from shareholders.

Brusuelas thinks productivity is the more immediate issue. U.S. productivity growth had lingered near 1.5% before climbing toward 2.3%. Much of that increase, he says, came from pandemic labor shortages that forced companies to adopt better technology. ChatGPT-style tools are still too new to dominate the official data.

Still, AI investment is already large enough to affect the data. Brusuelas says AI-related equipment spending helped drive first-quarter GDP growth, though not all of it.

A country can become richer while many people become worse off. GDP can, at least theoretically, rise while paychecks get harder to find. Profits can climb while hiring slows. The stock market can reward the same changes that make families feel less secure.

AI may not cause the next bust. By the traditional framework, it’s more likely to create a boom. But if GDP keeps rising while millions of people struggle to find good work, it will only reinforce a growing distrust in official government pronouncements about the economy and metrics like GDP.