Reading at Advanced level
Despite Treasury Secretary Bessent and Fed Chair Warsh's assertions that AI-driven productivity will deflate away structural fiscal concerns, labor economists flag a troubling disconnect: workers' income share has reached 52.8% of GDP, the lowest since records began in 1947, while corporate profit margins expanded to a record 14.9% of output. Gregory Daco (EY-Parthenon) argues this divergence reflects pre-AI secular trends—a decade of automation, post-pandemic cost discipline, and capital deepening—yet warns that 50% may not constitute a durable floor. "As long as you continue to see concentrated gains on the capital side, and within a certain number of firms," he told Fortune, labor's share could contract further. Q2 data illustrates the mechanism: real compensation growth flatlined (2.6% nominal compensation against underlying inflation), while 1.7% output growth derived from merely 0.3% additional labor input—a dramatic productivity pickup captured by margin expansion rather than wage acceleration.
The historical precedent—railroad and dotcom booms—suggests technological revolutions front-load capital gains to a narrow cohort of vertically integrated incumbents; wage mean-reversion typically lags by years. Yet the AI cycle exhibits unprecedented capital intensity. PwC estimates data center capex could reach $31 trillion by 2050—nearly coequal to current GDP—yet a structural leak undermines the domestic multiplier. Annualized imports of GPU servers (Census "large computers" category) surged to $450 billion last month, a ninefold increase from ~$50 billion through 2023. Since imported capex adds to investment while simultaneously subtracting as imports (net GDP contribution: zero), Hilsenrath and others observe that robust capital spending and productivity coexist with subdued hiring and compressed real wage growth. Manufacturing and construction have bucked recessionary headwinds solely through data center buildout; without it, Federal Reserve Beige Book commentary suggests contraction elsewhere would dominate.
The policy implication cuts sharply: if capital concentration persists and offshore capex sourcing limits domestic labor absorption, productivity gains may fail to propagate through traditional wage channels. Daco notes that prior technological diffusion (1990s software productivity) eventually broadened benefits; no equivalent guarantee inheres in an AI paradigm characterized by winner-take-most dynamics and historically compressed capex sourcing. Labor's structural share compression—absent policy intervention or demand-side shocks that restore wage bargaining power—may prove durable rather than cyclical.