AI boom and official optimism
Scott Bessent, the United States Treasury Secretary, and Kevin Warsh, the Federal Reserve Chairman, argue that the coming wave of artificial intelligence will make the country richer and even deflationary, reducing concerns over the $40 trillion debt load.
Workers' share of income falls
Analysts, however, are questioning who will benefit. The portion of national income going to labour has dropped to 52.8 percent, the lowest figure recorded since the government began tracking the metric in 1947. At the same time, corporate profit margins have risen to a record 14.9 percent of GDP.
"Productivity growth protects margins, not income," said Gregory Daco, chief economist at EY-Parthenon.
The role of productivity and automation
According to Daco, the productivity gains that underpin this divergence largely pre-date the AI boom. In the second quarter, economic output grew 1.7 percent while hours worked rose only 0.3 percent. Compensation increased 2.6 percent, which, after adjusting for oil-driven inflation, translates to a flat or slightly negative change in real terms.
"As long as you continue to see concentrated gains on the capital side, and within a certain number of firms," labour's share could keep plummeting, Daco added.
He notes that a decade of automation, tighter hiring after pandemic-era expansion, and capital spending have driven productivity, with AI so far adding further concentration rather than broad-based gains.
Capital intensity and import dynamics
The AI surge is uniquely capital-intensive. PricewaterhouseCoopers projects data-centre investment to reach $31 trillion by 2050, roughly the size of today's US GDP. Yet much of the equipment is imported. Net imports of "large computers", the Census category for GPU servers, surged to an annualised $450 billion last month, up from about $50 billion a year earlier. In GDP accounting, an imported server adds to investment and is subtracted as an import, leaving a net contribution of zero.
Policy questions ahead
While capital spending booms and margins expand, hiring remains weak, housing struggles under higher mortgage rates, and the labour share continues to shrink. Former Wall Street Journal Fed reporter Jon Hilsenrath observes that US investment is booming while gross domestic product growth stays modest.
The divergence raises a difficult choice for Warsh and Bessent: allow the AI-driven expansion to run its course or intervene to ensure broader income distribution. Growth alone does not guarantee that the benefits reach workers, especially if most of the upside accrues to owners of data-centre assets and shareholders.
Higher long-term rates increase borrowing costs, dampening mortgage markets and homebuilding, which could further limit the spread of prosperity. As Daco concludes, there is no clear floor for labour's share of income, and the trajectory will depend on how policymakers respond to the concentration of AI-related gains.