U.S. equity markets are at record levels, supported by resilient corporate earnings and investors looking past geopolitical risks in the Middle East. However, a widening gap is emerging between asset prices and underlying economic conditions.
May 7, 2026
By Tyler Frawley, CFA
The consumer remains central to the U.S. economy, accounting for nearly 70 percent of GDP, but forward indicators have deteriorated. University of Michigan sentiment readings have fallen to historic lows, inflation expectations have reaccelerated, and lower-to-middle-income households face pressure from higher energy costs and pandemic-era savings drawdown.
The personal savings rate has slipped to 3.6 percent, down from 5.1 percent last year and 6.2 percent in Dec. 2019—just prior to the pandemic. Excluding the 2022 pandemic distortion, this was the weakest reading since 2008—during the Great Recession. The implication is straightforward: household balance sheets have far less capacity to absorb incremental price shocks than they did just a few years ago.
For now, the labor market continues to provide support. Employment remains relatively firm, offsetting inflation pressure at the household level. Structurally, however, we are in a “low-hire, low-fire” environment—stable on the surface, increasingly stagnant underneath. This balance tends to be less durable if hiring momentum continues to slow.
Market performance remains highly concentrated. A small group of mega-cap technology and semiconductor companies have driven most of the S&P 500 gains in recent years, sitting at the center of a massive AI infrastructure investment cycle—one of the largest capital spending waves in history.
This wave has become something of a macro support pillar. In what could otherwise be considered a sluggish economy, AI-related capex acts as a key growth driver. The central question is what this spending ultimately produces.
In traditional cycles, capital investment expands capacity and supports job creation. This cycle is structurally different. Much of this AI buildout aims not for labor expansion but labor substitution—particularly across repeatable white-collar tasks.
That substitution creates a second-order effect. As these systems scale, demand for certain knowledge work categories declines. Early signals are already visible. White-collar job creation has slowed materially—year-over-year growth has been negative for 31 consecutive months, unprecedented in history outside recessionary periods. Broader employment growth has also stalled, with total U.S. jobs remaining roughly flat over the past year.
Note: Shaded areas indicate U.S. recessions.
Source – RBC Wealth Management, Bureau of Labor Statistics, Bloomberg; data through Q1 2026
The chart shows the year-over-year change in U.S. professional and business services (white collar) employment (in thousands) from 1976 through Q1 2026. The line fluctuates mostly between 0 and 1,000 for decades, with sharp drops during recessions including the early 1980s, 1991, 2001, and especially the 2008–2009 financial crisis. A deep plunge occurs around 2020, followed by a strong rebound. In recent years, the line has trended lower, falling below zero for 31 consecutive months as white collar job creation stagnates.
Recent workforce reductions from large technology firms reinforce this trend. In April, Meta announced layoffs affecting roughly 10 percent of its workforce—about 8,000 employees plus 6,000 open roles—citing AI-related cost pressures. Microsoft introduced a voluntary retirement program for the first time in its history, impacting roughly seven percent of its U.S. workforce to fund AI infrastructure. Perhaps the most telling signal, in our view, came in February, when Block (formerly called Square, a financial technology firm) announced it would lay off 40 percent of its approximately 10,000 employees, with AI explicitly earmarked as replacement. Across these examples, the common theme is clear: capital is being redirected toward AI, even at the expense of labor.
The key risk is diffusion. If efficiency programs remain confined to technology, the macro impact is likely limited. If they extend into finance, healthcare administration, law, logistics, and other services-heavy sectors, the employment impact becomes structural. Wage growth slows, hiring weakens and consumption naturally follows.
That is the core tension: the same tools driving profit margin expansion may gradually erode the employment base that supports demand and sustainable economic growth.
Some frameworks describe a more extreme outcome referred to as “ghost GDP.” In this scenario, AI-driven productivity and corporate profits remain strong while employment and income lag. As AI displaces high-earning workers, consumption weakens even as output rises, creating a loop where machines generate GDP growth but do not spend or consume. The result is stronger headline growth masking income erosion, inequality and underlying demand fragility.
It’s important to note that this is not the base case for RBC economists, though we believe it serves as a directionally important thought experiment. If AI-driven productivity meaningfully slows white-collar employment growth, the transmission mechanism into consumption becomes a central macro risk. Given that consumption dominates U.S. GDP, even modest labor income weakness could have outsized effects.
The counterargument is well established: technological revolutions destroy some jobs but ultimately create more over time. The internet expansion of the 1990s is the most cited parallel, delivering both productivity gains and strong employment growth, particularly in white-collar sectors.
However, the current cycle looks different. Despite unprecedented capital spending on AI infrastructure, labor demand—especially in knowledge work—has not strengthened but, in many areas, has softened. This raises a key question: why haven’t these massive investments translated into stronger job creation?
The distinction versus the 1990s is functional. The internet expansion required broad human adoption, integration and system buildout across industries. AI is more directly substitutional. It automates tasks that previously required skilled human input, rather than expanding entirely new labor categories at the same pace.
That difference matters. If AI reduces the need for certain roles faster than new roles are created, the adjustment becomes more compressed and potentially more disruptive than prior cycles.
U.S. equity markets are operating in a narrow equilibrium. On one side, AI-driven investment and margin expansion support valuations and mask softness in parts of the economy. Conversely, those same forces may gradually weaken the labor income base that supports long-run consumption.
This creates a divergence that is difficult to sustain indefinitely. Markets are pricing strong productivity and earnings power, while the real economy shows slower labor momentum and weaker household balance sheets.
The long-term productivity case for AI remains intact. The more relevant question is how these efficiency gains are distributed across labor, income and demand.
If the efficiency cycle accelerates while household financial buffers continue to erode, the gap between market performance and underlying economic conditions will likely widen further. At some point, that divergence will need to be resolved.
Whether the economy can transition into a more efficiency-driven structure without meaningful adjustment in labor and consumption dynamics will likely be one of the defining macro questions for the rest of this decade.
RBC Wealth Management, a division of RBC Capital Markets, LLC, registered investment adviser and Member NYSE/FINRA/SIPC.
We want to talk about your financial future.
Investment and insurance products offered through RBC Wealth Management are not insured by the FDIC or any other federal government agency, are not deposits or other obligations of, or guaranteed by, a bank or any bank affiliate, and are subject to investment risks, including possible loss of the principal amount invested.