An independent tally reports around 128,500 tech job cuts across nearly 300 companies since January 2026, already exceeding the total for the whole of 2025. Over the first ten days of September, more than 6,300 cuts were announced, including around 3,300 at a single operator, roughly 10% of its head-office workforce. Reallocation towards AI is regularly cited as the reason.
We have covered job cuts explicitly justified by AI and, conversely, a survey in which nine out of ten executives see no effect. These two sets of facts coexist, and we need to explain how.
What the figures establish
Three things are solid.
The volume is real and rising. Exceeding a full year's total in eight months is not an optical illusion.
So is the sectoral concentration. These cuts hit technology massively, not the economy as a whole.
AI is cited. Corporate communications regularly mention a reallocation of resources towards AI to justify these decisions.
What the figures do not allow us to conclude
This is where rigour is required, because it is the point where all commentary goes off the rails.
The stated reason is not the cause. A company cutting jobs chooses its wording. Invoking a reallocation towards AI is more presentable than admitting overstaffing, a commercial slowdown or a strategic error. We put forward this hypothesis in August: AI serves as an acceptable justification for decisions taken for other reasons.
The sector is correcting a massive hiring period. Tech headcounts grew at an exceptional rate between 2020 and 2022. Some of these cuts reflect a return to the mean, independently of any technology.
Interest rates weigh heavily. A more costly financing environment mechanically reduces tolerance for headcount and unprofitable projects. It is a powerful and thoroughly boring factor, and therefore rarely cited.
A detail reported this week deserves attention. A large technology company is reportedly asking individual employees in its AI department whether they wish to become managers again, reversing a flattening-of-hierarchy policy presented as central. In other words: the same organisation that cut jobs to fund AI is rehiring management to steer what it has built. This is not the replacement of human work by machines. It is a reorganisation, with its mistakes and its reversals.
The most likely mechanism
Pulling together what we know, a coherent explanation emerges, and it is less spectacular than the usual narrative.
Companies are not replacing employees with models. They are shifting budgets: what used to fund teams now funds compute, chips and specialist engineers. The effect on employment is real, and it does not operate through the automation of tasks.
This is consistent with what we reported on Sunday about a thousand engineers deployed at client sites: the bottleneck is human, and it creates jobs as much as it destroys them — elsewhere, and not for the same people.
What that does not change for those on the receiving end
It has to be said, because analytical nuance can become indecent.
Whether the cause is AI, interest rates or a post-growth correction changes absolutely nothing for someone who loses their job. The debate over attributing causes is useful for understanding and forecasting. It is not useful for the person receiving the letter.
What we wrote about the thesis of a different kind of transition remains the most solid point: the debate over the net number of jobs masks the more important one about the pace and who absorbs the shock.
What to take away
These figures are real, they are rising, and they do not prove what they are made to say.
The right question to ask anyone justifying job cuts by AI is the one we were already asking: what exactly are you basing this on? What measured gain, on what task, with what indicator. A survey covering thousands of companies finds almost no effect; it is legitimate to ask those invoking one to document it.