US Job Advertisements for "AI-Highly Impacted" Positions Have Increased by 46% Since Early 2021
Recent data released by the well-known macro analysis platform the kobeissi letter (founded by Adam Kobeissi) shows that job advertisements for "AI-highly impacted" positions in the US have increased by 46% since early 2021. This increase is significantly higher than the overall national job advertisement salary increase of 39%, and the 25% increase for "AI-lowly impacted" positions.
The data categorizes positions based on the degree of AI impact into three tiers: salaries for AI-highly impacted positions have risen by 46%; salaries for AI-moderately impacted positions have risen by 41%; the overall salary increase for all job postings in the US is 39%; and salaries for AI-lowly impacted positions have only risen by 25%. The salary gap between these three tiers has begun to widen significantly since mid-2025 and has continued to expand this year. The platform interprets this divergence as employers competing more fiercely for talent in AI-related positions, leading to a restructuring of the overall compensation structure in the US due to AI.
PwC's "Global AI Employment Barometer" report, set to be released in June 2026, provides consistent independent data: after analyzing over 1 billion job advertisements across 27 countries, the report found that positions requiring AI skills have a salary premium of 62%, up from 57% the previous year; companies highly impacted by AI have an overall salary increase of 24%, compared to 17% for companies with minimal AI impact; the number of AI-related positions has grown at a rate of 69%, approximately eight times the overall employment market growth rate of 9%; and entry-level AI-related positions have increased by 35% since 2019, while other entry-level positions have shrunk by 10% during the same period.
However, not all studies support this conclusion. A team led by Torsten Slok, chief economist at Apollo Global Management, found through comparisons of the Anthropic Economic Index and US Labor Department wage data that after 2023, the actual salary growth rate for high AI exposure positions (including computer programmers, customer service representatives, and financial analysts) is 6.7 percentage points lower than that for low exposure positions, resulting in an overall annual salary loss of approximately $28 billion for related employees. About 5.8 million American workers (3.7% of the workforce) are currently engaged in such high exposure positions; JPMorgan strategist Phil Camporeale interprets this as employees in related positions being "more concerned about losing their jobs, and therefore not demanding raises as they did in the past."
Research from the Federal Reserve Bank of Chicago presents a more moderate conclusion: between 2019 and 2024, employment in the most AI-impacted positions grew by 4.05%, while positions less impacted saw layoffs. However, during the same period, salaries across various positions generally increased, and the degree of AI exposure did not create systematic distinctions in salary growth; the study suggests that the AI exposure metric is better viewed as a signal of "occupational restructuring" rather than a direct indicator of job loss or salary increases. Similarly, a study from Yale University's Budget Lab found that 33 months after the release of ChatGPT, no significant labor market impact has been observed.
This debate surrounding the relationship between AI and salaries essentially reflects differing conclusions drawn from various data sources and statistical measures: studies focusing on job advertisements and overall company salary levels (such as the data cited by the kobeissi letter and the PwC report) are more likely to capture the salary uplift effects of newly created AI-related positions and skill premiums, benefiting job seekers with AI-related skills and strong bargaining power, as well as employers actively recruiting AI talent; in contrast, research focusing on the actual salary growth of specific existing positions (such as Apollo's analysis) captures the suppressive effects of AI automation on the bargaining power of workers in certain repetitive, replaceable roles, with the pressured group being employees in industries with deep AI penetration who perform partially replaceable work and lack bargaining power.
It is noteworthy that these two seemingly contradictory research conclusions are not mutually exclusive—they likely describe two sides of the same phenomenon: the salary premium for newly created AI-related positions and top AI-skilled talent continues to expand, while traditional employees in positions that can be partially replaced by AI face compressed actual salary growth. Both forces are occurring simultaneously, contributing to the current "polarization" of the salary landscape in the US labor market.
Source: Public Information
ABAB AI Insight
Similar technology-driven salary differentiation has occurred historically. During the 1990s to early 2000s, the proliferation of internet technology saw significant salary increases for positions with programming and data processing skills, creating a phenomenon known as the "digital divide" in salary differentiation; an earlier analogy is the wave of computerization in the 1980s, where economists like David Autor proposed the "skill-biased technological change" theory—each round of general technological revolution first creates salary premiums among a small group of individuals mastering new technologies, which then gradually spreads or replaces a broader labor force. The narrative represented by the data from the kobeissi letter and the PwC report that "AI-exposed positions have faster salary increases" is essentially the latest version of this historical pattern in the AI cycle.
The paths of capital and talent competition are clearly visible—PwC data shows that the growth rate of AI-related positions is eight times that of the overall job market, and entry-level AI positions have increased by 35% since 2019, while other entry-level positions have shrunk by 10%, indicating that companies are concentrating their recruitment budgets and training resources on AI-related skilled talent, creating a "winner-takes-all" talent competition; meanwhile, Apollo's research reveals another side of the funding path—companies are choosing not to raise salaries to retain employees whose skills can be partially replaced by AI and who lack bargaining power (such as customer service representatives and junior programmers), but instead are using their "greater fear of losing their jobs" to suppress salary increases, reallocating budgets that could have been used for general salary increases to the battle for scarce AI talent.
The current stage of the labor market is highly similar to the skill premium differentiation period before and after the bursting of the internet bubble in the early 2000s—during that time, salaries for those with web development and database management skills soared, while salaries for traditional office administrative positions stagnated. The difference is that the speed of this AI-driven differentiation is significantly faster: the PwC report shows that the AI skill salary premium rose from 57% to 62% within a year, a pace that far exceeds that of historical technological revolution cycles, indicating that the labor market is in an accelerated phase of "skill repricing," rather than a gradual adjustment phase.
This essentially represents a combination of technological substitution and the transfer of pricing power—on one hand, the labor force with AI-related skills has gained significantly stronger bargaining power, with their salary pricing power shifting from "overall labor market supply and demand" to "scarce AI skill supply and demand"; on the other hand, traditional workers in positions that can be partially replaced by AI have seen their bargaining power systematically weakened, with salary growth suppressed. Mechanistically, the reason this dual differentiation can occur simultaneously is that AI technology has fundamentally opposite effects on different positions: for a small number of high-skill positions, AI expands individual labor productivity and irreplaceability, thereby enhancing salary pricing power; for a large number of medium- to low-skill positions with standardizable tasks, AI directly reduces employers' dependence on workers in those positions, weakening the latter's negotiating leverage. These two forces acting simultaneously in the same labor market are the fundamental reason behind the seemingly contradictory yet coexisting conclusions of "AI-exposed positions having faster salary increases" and "AI-exposed positions experiencing actual salary pressure."
ABAB News · Cognitive Law
One technological revolution can simultaneously create winners and losers.
Only scarce skills can be priced; replaceable skills can only be bargained.
Those who receive salary increases are irreplaceable individuals, while those whose salaries are suppressed are in replaceable positions.