The share of national income going to American workers has slipped to its lowest point since the government began tracking the metric in 1947, a development that arrives well before the full force of artificial intelligence is felt across the economy. While Treasury Secretary Scott Bessent and Federal Reserve Chair Kevin Warsh herald an impending AIโdriven productivity surge that could render the nationโs $40 trillion debt manageable, analysts are asking who will actually reap the rewards. The divergence between soaring corporate profit margins and stagnant worker compensation is not a new phenomenon; it reflects a decadeโlong trend in which automation, capital spending, and cost discipline have boosted output without translating into broadโbased wage gains. Understanding this backdrop is essential for anyone trying to gauge whether the AI revolution will finally lift all boats or simply deepen existing inequities.
History offers a cautionary lens through which to view todayโs dynamics. In past technological wavesโsuch as the railroad expansion of the late 1800s and the dotโcom boom of the 1990sโearly gains tended to concentrate in the hands of a few large, vertically integrated firms that could afford the upfront capital investments. Smaller competitors faced persistent cost pressures, policy uncertainty, and higher financing costs, which limited their ability to share productivity improvements with employees. Over time, as the underlying technologies diffused and became cheaper, wage growth eventually followed, but the lag could stretch for years or even decades. The current AI environment mirrors this pattern: massive dataโcenter builds and GPU server purchases are being frontโloaded by a handful of tech giants, while the broader economy waits for the technology to become widely accessible and affordable.
Recent quarterly data illustrates the immediate consequences of this concentration. Economic output expanded by 1.7% in the second quarter, yet the increase in hours worked was a mere 0.3%, indicating that most of the growth came from productivity gains rather than additional labor. Nominal compensation rose 2.6%, but when adjusted for oilโdriven inflation that peaked during the spring and summer, the real change in workersโ pay ranged from flat to slightly negative. This mismatch between output and compensation is a classic symptom of capitalโbiased technological change, where machines and software enable firms to produce more with fewer workers, boosting profits while leaving wages stagnant.
Corporate profit margins have climbed to a record 14.9% of GDP, underscoring how efficiently firms are converting revenue into bottomโline earnings. At the same time, laborโs share of income has fallen to 52.8%, the lowest level on record. Gregory Daco, chief economist at EYโParthenon, emphasizes that the productivity gains driving this divergence largely predate the AI boom; they stem from a decade of automation, postโpandemic hiring pullbacks, and sustained capital expenditures. As long as the gains remain concentrated within a limited set of firms and accrue primarily to capital owners, there is no inherent floor preventing laborโs share from slipping furtherโpotentially toward the 50% mark or even lower.
The AIโspecific narrative adds a layer of intensity to an already capitalโintensive trend. Dataโcenter construction, GPU server imports, and related infrastructure spending are projected to reach $31 trillion by 2050, a figure rivaling the size of todayโs entire U.S. economy. While this spending signals robust investment appetite, much of the equipmentโparticularly the highโperformance chips and servers that power AI modelsโis manufactured overseas. Consequently, the surge in imports of โlarge computersโ (the Census classification for GPU servers) has hit an annualized pace of $450โฏbillion, a nineโfold increase from the roughly $50โฏbillion yearly average seen through 2023. In GDP accounting, each imported server adds to investment but subtracts an equal amount as an import, resulting in a net contribution of zero to domestic output.
This accounting nuance helps explain why headline GDP growth has remained modest despite the apparent boom in AIโrelated capital spending. Investment figures look strong on paper, but the offsetting import drag means that the domestic economy does not capture the full productive potential of the new hardware. The situation creates a paradox: firms are spending hundreds of billions on equipment that should eventually allow them to produce far more with less, yet a substantial share of that spending leaks abroad, limiting the immediate boost to domestic jobs and wages.
The ripple effects of this capitalโfocused expansion are visible in labor markets and the housing sector. Hiring has remained weak even as firms report strong profitability, reflecting a reluctance to add headcount when productivity gains can be achieved through automation. Meanwhile, tighter monetary policyโprompted in part by concerns over overheating in AIโrelated investmentโhas pushed longโterm interest rates higher, making mortgages more expensive and suppressing homebuilding activity. A Chicagoโbased manager quoted in the Federal Reserveโs Beige Book noted that construction and manufacturing are currently propped up by dataโcenter projects; without this niche demand, those industries would likely be in recession.
From a fiscal perspective, the shifting income distribution poses challenges for policymakers who rely on a broad tax base to fund government services. If an everโlarger slice of national income flows to capital owners and shareholders, the pool of taxable wagesโhistorically the most stable and progressive source of revenueโshrinks. Simultaneously, the political constituency that traditionally advocates for proโgrowth, proโworker policies may diminish, making it harder to build consensus around measures such as workforce training, wage subsidies, or progressive taxation. The risk is a feedback loop where slower wage growth undermines consumer spending, which in turn tempers overall economic dynamism.
The capital intensity of the AI buildโup also creates competition for finite financial resources, exerting upward pressure on borrowing costs across the economy. Hundreds of billions of dollars earmarked for AI infrastructure vie for the same pool of savings that could otherwise fund smallโbusiness expansion, affordable housing, or greenโenergy projects. As longโterm rates climb, the cost of financing everything from corporate bonds to consumer mortgages rises, potentially dampening interestโrateโsensitive sectors. The Wall Street Journal highlighted this dynamic with a chart showing how higher longโterm rates correlate with declining homeโstarts, underscoring the tradeโoff between financing AI expansion and maintaining broadโbased economic vitality.
None of this suggests that the AI productivity boom will fail to deliver genuine economic gains; rather, it highlights that the distribution of those gains is not automatic. History shows that the eventual diffusion of transformative technologies can lift wages, but only after a period of adjustment, policy intervention, and broadโbased access to the new tools. For workers, the immediate priority is to acquire skills that complement AIโsuch as data interpretation, AIโmodel oversight, and creative problemโsolvingโso they remain indispensable in a more automated landscape. For investors, diversification into sectors that stand to benefit from AIโenabled efficiency gains, while also hedging against potential overconcentration in a few megaโcaps, may mitigate risk. For policymakers, consider targeted incentives that encourage domestic production of AI hardware, expand access to broadband and computing resources for small businesses, and strengthen wageโsupport mechanisms such as earnedโincome tax credits or sectorโspecific training grants. By acting now to shape how AIโs productivity dividends are shared, there is still a chance to prevent the labor share from slipping past any conceivable floor.