Most layoffs are described in headcount, because headcount is abstract and cheap to say. Occasionally a company is forced by accounting rules to describe one in dollars instead, and the effect is bracing. On July 29, Meta filed its second-quarter results and put a number on the roughly 8,000 people it removed in May: $1.18 billion in severance.
Divide it out. That is about $147,500 per person, booked as a one-time charge, in a quarter where revenue rose 28% to $60.8 billion. Nobody at Meta had to argue that the company couldn’t afford these people. It could. It chose not to want them, and then wrote a check large enough to make the choice show up on the income statement.
The layoff cost more than it saved this quarter
Net income for the quarter came in at $15.85 billion — down 14% year-over-year, in a quarter with record revenue. The two culprits were a $2.4 billion legal charge and that $1.18 billion of severance. Strip both out and Meta’s operating income would have been up about 9% instead.
Read that sequence slowly. The stated logic of an AI-era layoff is efficiency: fewer people, lower cost, more margin. In the quarter it actually happened, Meta’s cut made the margin worse by more than a billion dollars. Efficiency arrives later, in theory, in a future quarter, on a slide. The bill arrives now, in cash, and it is specific.
This is the part the “AI made us do it” genre never covers. Firing people is not free and it is not fast. It is a procurement decision with an invoice attached, and the invoice at Meta’s scale runs to roughly the price of a mid-sized acquisition. When a company signs that invoice anyway, it is telling you it believes the arithmetic works out over several years — which is a much stronger claim than “we are trimming.”
Headcount barely moved, which is the actual story
Here is the detail that reframes everything: after removing about 8,000 people, Meta ended the quarter with 75,472 employees, down just 1% year-over-year.
Eight thousand exits. One percent shrinkage. Those numbers only reconcile one way — Meta was hiring hard the entire time, into different roles. This was never a downsizing. It was a substitution, executed at speed and paid for at premium, in which one population of employees was removed and another was brought in behind them.
That distinction matters enormously if you work there, or anywhere like it. A downsizing is a weather event: the company is smaller, the storm passes, survivors stay put. A substitution is structural. The company is the same size next year, and possibly bigger, and your role is still gone — because the question was never “can we afford this headcount,” it was “is this the headcount we want.” Meta answered the second question with $1.18 billion, and Mark Zuckerberg spent the earnings call talking about self-improving systems and shipping agent products by October.
The uncomfortable read
Meta also narrowed its 2026 capital-expenditure guidance upward, to $130–145 billion. Set that against $1.18 billion of severance and the ratio is almost comic: the cost of removing 8,000 people is roughly one percent of what Meta plans to spend on infrastructure this year. Severance is a rounding error against the buildout. It will never be the thing that stops a cut.
So the protective instinct that says “I’m too expensive to fire” has it exactly backwards. Meta demonstrated it will pay $147,500 a head, take a 14% hit to quarterly net income, and absorb the reputational noise, to get the workforce composition it wants. Cost is not the constraint. Fit is.
The only durable position is being on the list of roles a company is hiring into during its own layoff — because that list existed at Meta in May, and it is why the headcount line barely moved. Eight thousand people were on the wrong side of a swap that, from the outside, looked like a contraction. It wasn’t. The doors were open the whole time. They were just labelled differently.
