The Silent Squeeze: Is AI Quietly Depressing American Wages Before It Claims Jobs?

As the U.S. labor market navigates the post-pandemic era, a new, unsettling narrative is emerging from the economic data. While job growth in August 2026 managed to surpass analyst expectations, the celebratory tone is being tempered by a persistent, nagging reality: wage growth is stalling, and inflation is once again outpacing paychecks. This divergence has forced economists to confront a provocative and potentially volatile question—is artificial intelligence exerting downward pressure on worker compensation long before it renders those same roles obsolete?

The anxiety surrounding AI has historically focused on the “displacement” narrative—the fear of mass layoffs as algorithms and automated systems take over the desks of white-collar workers and the stations of blue-collar laborers. However, as the 2026 data continues to roll in, the focus is shifting. Experts are increasingly looking at the “wage compression” theory: the possibility that companies are using AI not to replace workers, but to dampen their bargaining power and suppress salary growth.

The Shrinking Piece of the Pie: A Historical Context

The latest nonfarm payrolls report is far from an outlier. The Bureau of Labor Statistics (BLS) has released a cascade of data suggesting a cooling labor market, with the Employment Cost Index revealing that inflation-adjusted wages and salaries actually dipped by 0.4% year-over-year through June.

Even more concerning is a long-term trend that suggests a structural shift in the American economy. According to the BLS productivity report, labor’s share of nonfarm business output—the slice of the economic pie that goes to workers in the form of wages and benefits—sank to 52.8% in the second quarter of 2026. This figure marks the lowest level recorded since the series began in 1947. While some analysts point to decades of automation as the culprit, there is a growing consensus that AI, with its unprecedented capability to handle complex cognitive tasks, may be accelerating this trend at a velocity never before seen.

Chronology of a Shifting Labor Market

To understand the current tension, one must look at the unique circumstances of the last five years.

• 2021–2022 (The Tight Market Era): Following the initial shocks of the COVID-19 pandemic, the U.S. experienced an anomalous labor environment. Demand for workers surged, giving employees significant leverage to demand higher wages, particularly in the professional services and tech sectors.
• 2023–2024 (The Normalization Phase): As the post-pandemic “sugar high” faded, hiring slowed. Tech companies, which had over-hired during the lockdown era, began significant rounds of layoffs. Simultaneously, lower-pay sectors like hospitality and healthcare became the primary drivers of job growth, which mathematically dragged down the national average for wage gains.
• 2025–2026 (The AI Integration Phase): By mid-2026, AI tools moved from novelty to standard utility. Companies began integrating large language models and predictive analytics into core business functions. It is within this period that the correlation between “high AI exposure” and “stagnant wage growth” began to appear in econometric studies.

Evidence from the Frontlines: The Apollo Study

The most direct evidence for the wage-suppression theory comes from a recent study by Apollo Global Management’s chief economist, Torsten Slok, and his co-author, Sania Edlich. Their research isolated occupations classified as “highly exposed” to AI and compared them to those with minimal exposure.

The findings were striking: workers in high-exposure fields saw their real-wage growth stall, trailing behind their counterparts in less-exposed roles by a staggering 6.7 percentage points after 2023. Crucially, the study did not find a statistically significant spike in unemployment within these high-exposure groups. The implication? Firms appear to be capturing the efficiency gains provided by AI by holding wages flat, rather than by cutting headcount.

While the Apollo report is viewed as a bellwether for a new era of AI-impact measurement, labor experts urge caution. Ben Zipperer, a senior economist at the Economic Policy Institute, notes that the sample size—using only 321 of the roughly 800 BLS occupational categories—is relatively small. He argues that the “AI exposure” metric can be misleading.

“If AI makes it cheaper to produce software, the money saved doesn’t just vanish,” Zipperer explains. “It gets reallocated elsewhere in the economy. If you see wage stagnation in coding, you might be seeing a redistribution of wealth rather than an absolute loss of value. The ‘exposed’ jobs look worse in isolation, but that doesn’t mean the total economic pie is shrinking.”

The Expert Consensus: A Nuanced Future

The debate among top-tier economists is less about whether AI will have an impact, and more about where that impact will be felt most acutely.

Daron Acemoglu, a professor of economics at MIT, believes that the current data is limited by the fact that AI adoption is still in its infancy. "AI models are not yet widely adopted for many tasks," Acemoglu observes. "Therefore, some of the displacement effects may be currently underestimated."

However, Acemoglu argues that the U.S. labor market’s relative flexibility—combined with a weak social safety net—means that the primary pain point will likely be wages, not employment. "If the U.S. continues to develop AI primarily as an automation technology—under the goal of replacing human tasks—the wage impact will be significant," he warns.

The Expertise Paradox

David Autor, the head of the MIT economics department, offers a counter-narrative to the idea that "exposure" to AI is inherently negative. His research on accounting clerks and inventory clerks shows that two similarly "exposed" roles can have drastically different outcomes based on how the technology interacts with human expertise.

Accounting clerks, despite a 32% drop in employment over 40 years, saw their wages climb by 39% as they transitioned into more specialized, high-value roles. Inventory clerks, conversely, saw their employment numbers explode by 175% while their wages plummeted by 13%. The lesson is clear: AI does not treat all professions the same. It creates winners and losers based on whether the technology complements or replaces the core expertise of the worker.

The "Experience Premium" Warning

A critical concern raised by the Dallas Fed is the impact on the next generation of workers. Their research indicates that for entry-level roles, AI is already acting as a substitute for human learning.

In the traditional white-collar model, young professionals learn "tacit knowledge" by performing routine, codifiable tasks. If AI takes over those entry-level tasks, the "job ladder" is broken. Companies may find it cost-ineffective to train novices when an AI can do the junior work more efficiently. This creates a long-term sustainability crisis: if the entry-level rung is removed from the ladder, how does the workforce of the future develop the skills required to become experienced workers?

Implications: Moving Toward "Pro-Worker AI"

As the debate matures, the framing is shifting away from an adversarial "Human vs. Machine" conflict. Economists and policy experts are increasingly advocating for a "pro-worker AI" framework—an approach where technology is designed to enhance human capabilities rather than merely automate them out of existence.

Jennifer Huddleston, a senior fellow at the Cato Institute, suggests that policy should focus on education and adaptability rather than protectionism. "We are seeing the early stages of a transition," she notes. "The focus should be on how we enable workers to utilize these tools to create new categories of jobs and entrepreneurial opportunities."

The path forward remains uncertain. The 2026 data serves as a stark reminder that the impact of technology on labor is rarely a clean story of mass obsolescence. Instead, it is a slow, grinding adjustment of value, bargaining power, and the nature of work itself. Whether AI becomes a tool that lifts all boats or a wedge that drives a deeper divide in the American middle class will depend not just on the code, but on the policy frameworks and corporate strategies that define the next decade of American labor.