Challenger, Gray & Christmas, the Chicago outplacement firm whose monthly tally is the longest-running count of announced US redundancies, recorded 33,429 job cuts in July in a report released on 6 August, down 27% from June's 45,849 and the lowest monthly figure in two years. Artificial intelligence led every stated reason for a fifth consecutive month at 10,970 cuts, 33% of the July total against 24% of the 112,713 announced since January. Technology led every sector again with 9,867.1
Employers under scrutiny were widely expected to stop volunteering AI as a cause. Instead the monthly total shrank and AI's share of it grew. Announcements are self-reported and a firm can name AI to look decisive rather than distressed, which is why one survey found 59% of employers overstating it. Discount the series as hard as you like and it remains the only evidence here that nobody can accuse of index-building, because the employers are describing their own decisions rather than a model's estimate of them.
The quarter's filings let the same question be asked of the four largest AI spenders directly. Microsoft's annual report, filed on 29 July, records approximately 223,000 people employed on a full-time basis at 30 June, down from 228,000 twelve months earlier.2 Meta reported 75,472, down 1% year on year, a figure that still contains roughly 8,000 people affected by its May reduction who drop out of the count next quarter.3 Amazon reported 1,595,000, up 3%, on a line that counts full-time and part-time staff and excludes contractors, so it is not a full-time equivalent measure at all.4 Alphabet had already disclosed 198,933 staff, up 11,830 year on year.
Two of the four shrank and two grew, in the same quarter, in the same industry, with all four spending heavily on the same technology. Headcount direction does not track AI investment, and the definitions do not even agree well enough to make the comparison clean. That is why attribution keeps falling back on what companies say about their own reasons rather than on what they file.
