The numbers come first. Workers' share of U.S. income fell to 52.8% last quarter, the lowest level since the government began tracking the figure in 1947. Corporate profit margins, meanwhile, hit a record 14.9% of GDP.
Treasury Secretary Scott Bessent and Federal Reserve Chairman Kevin Warsh have told markets the AI productivity boom will soon make America richer — richer enough, Warsh has suggested, to stop worrying about the nation's $40 trillion debt. Analysts are now asking a sharper question: richer for whom?
Gregory Daco, chief economist at EY-Parthenon, says the productivity gains behind today's divergence mostly predate AI. 'Productivity growth protects margins, not income,' he wrote in a note last week. In the second quarter, economic output grew 1.7% on just 0.3% more hours worked. Compensation rose 2.6%, but against a spring and summer of oil-driven inflation, that comes out to flat or slightly negative in real terms, Daco told Fortune.
'I don't think there's a floor,' Daco said of the labor share, adding that as long as capital gains stay concentrated in a small number of firms, workers' cut could keep shrinking.
History offers a warning. In the railroad boom of the 1800s and the dot-com revolution of the 1990s, large, vertically integrated firms captured early gains while smaller competitors absorbed cost pressures and policy uncertainty. In the '90s, cheaper software eventually spread productivity — and wages — across the economy. Daco notes there is no guarantee AI follows the same script; this buildout is unusually capital-intensive, with data-center investment projected to hit $31 trillion by 2050, according to PricewaterhouseCoopers — nearly the size of current U.S. GDP.
A Chicago manufacturer told the Federal Reserve's Beige Book this week that construction and manufacturing are only holding up because of data-center spending; without it, the sector would be in recession.
Much of that equipment, however, is not American-made. Net imports of the large computer servers used for AI hit a $450 billion annualized pace last month, up from roughly $50 billion a year through 2023, per Census data compiled by economist Joseph Politano. Because an imported server adds to investment and subtracts as an import in equal measure, its net contribution to GDP is zero. Former Wall Street Journal Fed reporter Jon Hilsenrath, now at Serpa Pinto Advisory, put it plainly: 'While U.S. investment is booming, growth in gross domestic product has been modest.'
Capital rewards clear rules, and so far the AI buildout is rewarding the owners of capital first. That is not, on its own, a market failure — it is what happens when a technology arrives that is capital-intensive by nature. But if the gains keep accruing to a handful of data-center owners and their shareholders while the tax base and the workforce that fund a $40 trillion debt grow only slowly, Washington's bet that AI will simply solve the fiscal math looks premature.
The better answer is not redistribution dressed up as industrial policy, but more competition: lower barriers to entry, faster permitting for the smaller firms squeezed by 'persistent policy uncertainty,' and a Federal Reserve that lets productivity, not political promises, set the pace. The market has already voted on where AI's early profits are going. Whether they eventually reach the paycheck is still an open question — and one Washington cannot legislate away.
