The numbers have a momentum that is hard to argue with. Employers announced 97,006 job cuts in May, the highest May total since 2020, and according to Challenger, Gray and Christmas, AI was cited as the leading reason for the third month running, accounting for 38,579 of them, about 40 percent of the total. That share has climbed from 7 percent in January to 25 percent in March to 40 percent in May. A Mercer survey of 825 chief executives found that 99 percent expect AI to drive at least some headcount reduction within two years. On the surface, the long-forecast disruption has arrived on schedule.
Look closer at the verb, though. Challenger does not measure jobs that AI destroyed; it measures the reason companies give. And as the writer Bernard Marr argued this week, the explanation is suspiciously tidy. The same firms blaming AI for cuts are reporting record profits, buying back billions in stock, and committing hundreds of billions of dollars to chips, data centres and cloud capacity. Meta is cutting 20,000 roles while hiring aggressively for AI talent; Salesforce announced cuts while making AI acquisitions. Redirecting payroll into infrastructure is a strategic choice. Calling it automation is a more flattering story than admitting you are reshuffling the budget, and it spares executives the awkward language of ordinary cost-cutting.
Both things can be true. There is real evidence that AI is already reshaping specific corners of the labour market rather than the whole of it. An analysis applying an AI-substitution index to US employment found that jobs at high risk, such as graphic designers and economists, fell more than 4 percent between 2019 and 2025, while low-risk jobs like electricians and teachers grew 13 percent. The damage is concentrated, not yet broad, and it falls hardest on the junior and entry-level tasks that have always been the rungs people climb to build a career.
That concentration is what makes the public mood so combustible. A Center Square poll found roughly 70 percent of Americans at least somewhat concerned that AI could take their job, with anxiety highest among those with less education and lower incomes, exactly the groups who have least slack to absorb a shock. The short-seller Carson Block went further on a podcast this week, arguing it is "entirely conceivable" that 15 percent of knowledge workers could be displaced within a few years, and sketching a chain reaction in which laid-off workers drain retirement accounts, stock inflows reverse, and universal basic income stops being a fringe idea.
The strangest twist is that the labour market is short of workers, just not the ones being laid off. As the Financial Times noted, the same tech giants planning to spend 725 billion dollars on infrastructure are discovering they cannot find enough people to build it. The construction trades face a shortfall projected to top a million workers by 2030, and the fastest-growing US occupations are physical ones: wind-turbine technicians and solar installers. Writing in Foreign Affairs, scholars revisited the failed Trade Adjustment Assistance programme to warn that the right response is not arguing over whether AI counts as the cause, but building support fast enough and broadly enough to match the scale of disruption, whatever its label. On that, at least, the data and the skeptics agree: the institutions meant to catch displaced workers do not yet exist.