Did AI Fire You, or Was It the Alibi? Auditing This Year's 140,000 Layoffs

OUR THESIS. AI did not cause most of the 140,000 layoffs attributed to it this year: it is the alibi that makes them presentable to Wall Street. The hard data says two things at once. First: there is still no demonstrable mass replacement of workers by AI; a large share of these cuts stems from high interest rates, pandemic-era overstaffing and the need to fund gigantic data-center investments. Second: where there is a real, measurable signal is in junior employment, which is collapsing not through layoffs but through a hiring freeze. Confusing the two —the corporate excuse and the real damage— is the most widespread analytical error of the year.

THE ACCOUNTING THAT DOESN'T ADD UP. Start with the numbers. Challenger, Gray & Christmas —the historical reference for tracking U.S. job cuts— attributes 101,743 layoffs to AI between January and June 2026 alone, nearly double all of 2025 (54,836). Adding the following months, the cumulative figure since 2023 hovers around 150,000. Here is the first anomaly: the share of layoffs companies blamed on AI jumped from about 7% in January to roughly 40% by May, per that same tracker. Model capability did not improve fivefold in four months; narrative convenience did. Even Sam Altman, OpenAI's chief, has admitted there is “AI washing” — companies blaming AI for layoffs they would have made anyway. The market isn't fully buying it either: according to a Financial Times analysis cited by TechCrunch, companies that named AI in their cuts underperformed the Nasdaq by almost 10% in the 30 trading days after the announcement.

WHAT THE SERIOUS STUDIES SAY. Against the earnings calls —the quarterly results presentations aimed at investors— the academic evidence paints a far more modest picture. METR's randomized controlled trial (the gold standard: tasks randomly assigned with and without AI) found in 2025 that veteran developers were 19% slower using AI assistants on complex code, even though they believed they were 20% faster; METR's 2026 update now estimates a real gain of about 18%, a sign the tools are maturing — from a humble baseline. MIT economist Daron Acemoglu calculates that only about 5% of the economy's tasks can be profitably automated within ten years, with a total productivity gain below 0.7%. Yale's Budget Lab, after 33 months of data since ChatGPT launched, finds “no discernible disruption” in aggregate employment. And the New York Fed found that only 1% of surveyed service firms had actually laid anyone off because of AI. Against this evidence, “AI does the work of 100 people” is a marketing claim aimed at investors, not a labor statistic.

THE REASONS THAT DON'T MAKE THE PRESS RELEASE. If it isn't (only) AI, what is it? Three verifiable things. One: money stopped being free; interest rates went from 0% to over 5% between 2022 and 2023, and speculative hiring died with them. Two: the pandemic hangover; analysts quoted in the business press estimate Big Tech ended up overstaffed by 25% to 75%. Three: capex. Alphabet, Microsoft, Meta and Amazon will spend close to $700 billion on AI infrastructure in 2026, and cutting payroll while announcing share buybacks is how you square that bill while looking good to Wall Street. MIT professor Paul Osterman sums it up in Fortune: AI is “a perfect excuse” because “it makes it seem as if it's not our decision.” The protagonists give themselves away: Andy Jassy said Amazon's 14,000 layoffs were “not really financially driven, and not even really AI driven — it's culture,” while Monday.com's founders —who cut 20% of staff, about 620 people, while growing revenue 24%— insisted the move was “not made to reduce costs or replace people with AI.” The alibi and its denial, in the same statement.

THE REHIRING BOOMERANG. The clearest proof the replacement story was oversold is that it is being reversed. Forrester predicts half of all AI-attributed layoffs will be undone “in some form” by the end of 2026. Surveys covered by Fast Company and Forbes suggest around half of companies that cut jobs for AI rehired within six months, and one in three spent more on restaffing than it saved. The canonical case is Klarna: in 2024 it boasted that its AI agent did the work of 700 customer-service staff; in 2025 CEO Sebastian Siemiatkowski admitted to Bloomberg that “we went too far — we focused too much on cost, and the result was lower quality,” and began hiring humans again. One important nuance: many rehires come back at lower pay or offshore. The reversal doesn't restore the old world; it makes it cheaper.

THE REAL SIGNAL: THE FIRST RUNG. And yet it would be a mistake to conclude “nothing is happening.” Something is — just elsewhere. The “Canaries in the Coal Mine” study by Erik Brynjolfsson and the Stanford Digital Economy Lab, built on real ADP payroll data, shows that employment for workers aged 22–25 in the most AI-exposed occupations is already 19% below where it would be had it kept pace with everyone else (August 2026 update). The mechanism is not layoffs: entry-level openings have simply vanished. Where AI substitutes for tasks, young employment falls; where it complements them, it holds or grows. This is the metric we will audit every quarter, because an economy that stops hiring juniors is burning through its future supply of seniors.

OUR READING AND THE IMPLICATIONS. Our reading is twofold and not contradictory. In the short term, the pain is real but mislabeled: most of these 140,000 layoffs are financial decisions dressed up as technological inevitability, and that label matters, because it absolves the executives doing the cutting and pollutes the regulatory debate with a panic the data does not yet support. The case-by-case audit —did total headcount actually fall? did output actually rise?— almost never accompanies the headline. In the long term we keep our tempered optimism: if AI ultimately delivers what it promises, the destination is not a workless economy but a more abundant one, where work shifts toward judgment, human relationships and what people are passionate about; the transition, as we keep saying, will be hard and uneven — and its first documented victim is the junior hire. Practical implications: demand from every “AI layoff” announcement the productivity evidence that almost never arrives; watch Stanford's dashboard as the most honest leading indicator we have; and, in public policy, attack the real problem —rebuilding the first career rung through training and entry-level hiring— instead of regulating the story told in press releases.