Lead U.S. economist Bernard Yaros of Oxford Economics recently said something that really resonated with me: the economy expanded last year “even though it was a very unfavorable backdrop.” He wasn’t being sarcastic. He was describing a truly bizarre year in which the economy continued to grow despite hiring slowing, inflation refusing to cooperate, and consumer sentiment sitting in the basement. Not dramatically, but enough to make people wonder what exactly is propelling American growth at the moment and whether anyone really knows the answer.
In summary, GDP increased by 2.2% in 2025. In contrast, Jobs failed to keep up. Economists have begun referring to this discrepancy between output and employment as a “jobless expansion,” although the term tends to make people uneasy for understandable reasons. It suggests prosperity without involvement. A banquet whose guest list is getting smaller.
| Category | Details |
|---|---|
| Topic | U.S. Economic Growth vs. Labor Market Slowdown |
| Key Metric: GDP Growth (2025) | 2.2% — down from 2.8% in 2024 |
| Key Research Source | Oxford Economics |
| AI Contribution to GDP | +0.4 percentage points (2025) |
| Consumer Spending Sensitivity | ~$0.05 increase per $1 of stock wealth gain |
| Net Immigration (2025 est.) | 160,000 — vs. 1.1 million in normal years |
| Labor Force Participation (Current) | 62% — projected to fall to 54% by 2050 under rapid AI adoption |
| Key Voices | Bernard Yaros (Oxford Economics), Martha Gimbel (Budget Lab at Yale), Federal Reserve Governor Michael Barr |
| Research Institutions | Fed Reserve Bank of Chicago, Stanford, Yale, University of Pennsylvania |
| Wealthiest 10% Projected Share by 2050 | 80% of total household wealth (rapid AI scenario) |
Go through the numbers, and you’ll see that they do tell a strange tale. Household wealth increased, mostly due to stock portfolios and home values rather than wages. Yaros discovered that this wealth directly influences spending, contributing about five cents to every dollar increase in equity markets. Americans spent as markets flourished, and the economy continued to expand as a result of that consumption. This may have performed better than anyone anticipated. It might also be a mechanism that fails to withstand a significant market correction.
Spending on AI, which is quiet and mostly unseen by the general public, seems to have had a significant impact. Investments in research infrastructure, data centers, and information processing equipment are expected to boost GDP growth by 0.4 percentage points in 2025. That number is not insignificant.
You can practically feel the weight of any large data center being constructed in Texas or Virginia from the concrete, cooling infrastructure, and fiber being trenched alongside former highway routes. The building is authentic. However, unlike a traditional manufacturing expansion, the jobs it creates are specialized and limited.
One aspect of this narrative that receives less attention than it merits is immigration. This year, net immigration is predicted to reach about 160,000, compared to a historical average of about 1.1 million. That is a structural decrease in the pool of available labor, not a marginal shift.
Research from the San Francisco Fed has shown that employment growth typically closely follows flows of immigrant workers. There will be fewer jobs created if there are fewer arrivals. The math is straightforward. Quieter borders would be sufficient to explain a slowing labor market without the need for AI.
Beneath all of this is the demographic reality. The number of older Americans retiring is outpacing their replacement. Some businesses are discreetly reducing their workforce as they continue to process the excessive hiring they made after the pandemic. Others have halted hiring because tariff uncertainty has made planning costly and challenging, not because of concerns about AI.
“There’s so much going on right now,” stated Martha Gimbel, executive director of Yale’s Budget Lab, recently, “and each of them feeds into the other.” That word kept coming up. It encapsulates an honest aspect of attempting to understand this economy: the causes are intertwined.
In a speech this week, Federal Reserve Governor Michael Barr hinted at the possibility of a real “jobless boom”—a situation in which the adoption of AI picks up speed to the point where employment declines while the economy grows. He was cautious, almost tactful. He pointed out that in the more gradual scenario, the economy stumbles through without a crisis while some jobs disappear and new ones appear. The image becomes significantly darker in the fast scenario.
According to a study conducted by researchers from the Chicago Fed, Stanford, Yale, and Penn, the labor force participation rate could drop from its current 62% to 54% by the middle of the century if artificial intelligence (AI) advances quickly and surpasses human performance in most cognitive tasks by 2030. Rather than aging or demographics, AI would be responsible for about ten million of those lost jobs.
The GDP forecast that goes along with that projection is what makes it worthwhile. Economists predict that by the late 2040s, annual growth will reach 3.5% under the same rapid scenario. Forecasts above 5% are made by AI experts, who are typically more optimistic.
That growth is remarkable. Additionally, it means that by 2050, the richest 10% of households could own 80% of all wealth. Growth without widespread participation is more than just a financial issue. It usually turns into a political one.
It’s still unclear if we are currently witnessing the first few chapters of that tale or if we are in a more typical cyclical period that will end on its own once immigration levels stabilize and businesses have finished processing their excess labor.
The macro data appears to be fairly stable. However, researchers monitoring workers in the most AI-exposed occupations between the ages of 22 and 25 have discovered a 13% relative employment decline—a figure that doesn’t appear in the headline figures but remains there, waiting. The economy expands. Who it grows for is the question.
